forked from swhan0329/codex-101
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathi18n.js
More file actions
1412 lines (1410 loc) · 196 KB
/
Copy pathi18n.js
File metadata and controls
1412 lines (1410 loc) · 196 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
// i18n translation data
const translations = {
ko: {
nav_guide: "가이드",
nav_use_cases: "활용 사례",
hero_badge_prefix: "공식 문서 기준 재검수",
hero_title: "OpenAI Codex를<br/>실무에 바로 붙이는 가이드",
hero_sub: "CLI · Desktop App (macOS/Windows) · IDE Extension · Web · Mobile · SDK automation<br/>최신 공식 문서를 기준으로 어디서 시작할지, 어떤 모델을 쓸지, 샌드박스/승인, MCP, 자동화를 처음 사용자와 실무 사용자 모두 읽기 쉽도록 다시 정리했습니다.",
hero_cta: "빠른 시작 경로 보기",
hero_cta_secondary: "Prompting 실전 가이드",
hero_jump_label: "챕터 바로가기",
hero_panel_kicker: "Launch paths",
hero_panel_title: "지금 필요한 흐름부터 바로 들어가기",
hero_panel_desc: "처음 실행, 팀 운영, 프롬프팅 실전 중 현재 목적에 맞는 경로를 먼저 잡고 아래 섹션으로 내려가면 훨씬 덜 헤맵니다.",
hero_path_1_tag: "First run",
hero_path_1_title: "10분 안에 설치와 첫 작업",
hero_path_1_desc: "앱 설치, 로그인, 첫 로컬 작업, Git 체크포인트까지 빠르게 훑습니다.",
hero_path_2_tag: "Team setup",
hero_path_2_title: "프로젝트 규칙과 도구 권한 정리",
hero_path_2_desc: "AGENTS.md, config.toml, MCP, 승인 방식을 팀 기준에 맞게 정합니다.",
hero_path_3_tag: "Prompting",
hero_path_3_title: "긴 작업이 덜 흔들리는 요청 방식",
hero_path_3_desc: "도구 사용, 결과 검증, 실행 설정을 한 번에 정리합니다.",
hero_signal_1: "chapters",
hero_signal_2: "실시간 언어 전환",
hero_signal_3: "공식 문서 재검수",
toc_title: "가이드 맵",
toc_lead: "먼저 상위 챕터로 전체 지형을 파악하고, 필요한 깊이는 섹션 17에서 바로 이어서 읽으세요.",
toc_home_1_t: "Foundations",
toc_home_1_d: "Codex가 무엇인지, 어떤 제품과 모델 구성이 있는지 먼저 잡습니다.",
toc_home_2_t: "Setup and Surfaces",
toc_home_2_d: "설치, 인증, CLI/App/IDE/Web 각각의 쓰임새를 빠르게 훑습니다.",
toc_home_3_t: "Execution Model",
toc_home_3_d: "승인 모드, 슬래시 명령, steering 방식처럼 실제 실행 흐름을 이해합니다.",
toc_home_4_t: "Context and Tools",
toc_home_4_d: "AGENTS.md, config.toml, MCP로 Codex의 행동 규칙과 도구 권한을 정리합니다.",
toc_home_5_t: "Sessions and Automation",
toc_home_5_d: "세션 운영, codex exec, 자동 리뷰, CI/CD 연계를 정리합니다.",
toc_home_6_t: "Prompting Codex Agents",
toc_home_6_d: "출력 형식, 도구 사용, 실행 설정, starter playbook을 한 섹션 안에서 바로 정리합니다.",
toc_home_7_t: "Decision Help",
toc_home_7_d: "고급 활용, FAQ, 공식 레퍼런스로 마무리합니다.",
home_nav_kicker: "Guide map",
home_nav_title: "Codex 101 chapters",
home_nav_lead: "최신 사실 확인부터 시작 방식 선택, 실행 규칙, 자동화, prompting까지 실무 적용 순서대로 묶었습니다.",
overview_kicker: "Start here",
overview_title: "처음 사용자와 실무 사용자를 위한 읽기 경로",
overview_lead: "이 페이지는 먼저 두 가지를 보여줍니다. 지금 어떤 모델과 사용 방식을 고르면 되는지, 그리고 처음 사용자와 실무 사용자가 어디서부터 읽으면 좋은지입니다. 오늘 반영한 변경 내용은 하단 Changelog에만 따로 모았습니다.",
overview_update_1_title: "모델/사용 방식 매트릭스 재확인",
overview_update_1_desc: "`codex/models`를 2026년 6월 3일 기준으로 가장 먼저 다시 확인했고, `gpt-5.5` 최우선 추천, `gpt-5.4` 대체 선택지, `gpt-5.4-mini`의 빠른 로컬/subagent 역할, `gpt-5.3-codex`의 Cloud/code review 역할, `gpt-5.3-codex-spark`의 Pro research preview 역할이 유지되는지 대조했습니다. 특히 오늘도 `codex/models`가 `gpt-5.5`의 API Access를 true로 표시하는 반면 `codex/pricing`의 API Key 표는 아직 Not available로 두는 차이를 분리해 설명했습니다.",
overview_update_2_title: "요금제·크레딧 구조 재정리",
overview_update_2_desc: "개인 플랜 한도뿐 아니라 `GPT-5.5` 사용량 표, `Free $0`·`Go $8`·`Plus $20`·`Pro from $100`, Plus의 web/CLI/IDE/iOS 접근, Business·Enterprise의 credits 설명, 그리고 현재 `codex/pricing` 문서의 Plus / Pro 5x / Pro 20x / Business / API Key 구성을 다시 맞췄습니다. 가격표에는 Sites가 Business/Enterprise에서 available, Plus/Pro/API Key에서는 unavailable로 표시되고, Sites FAQ는 preview 기간 동안 무료라고 설명합니다. API-key 사용자는 실제 Codex 모델 선택기와 표준 API pricing을 같이 확인해야 한다는 주의점을 가격 섹션에서 더 또렷하게 분리했습니다.",
overview_update_3_title: "Quickstart·IDE·Mobile 경로 정리",
overview_update_3_desc: "Quickstart 기준으로 처음 사용자는 App 중심, 실무 사용자는 IDE/CLI + Docs MCP 중심으로 읽히도록 시작 경로를 유지하되, remote connection 설명을 오늘 공식 문서에 맞춰 다시 잡았습니다. 공식 문서는 이제 ChatGPT iOS/Android 모바일 앱이 awake/online macOS 또는 Windows Codex App host를 제어할 수 있다고 설명합니다. 연결된 host의 파일, credential, permission, plugin, browser setup, Computer Use, local tool이 그대로 쓰이고, 같은 account/workspace와 필요한 SSO/MFA/passkey가 맞아야 합니다. Business/Enterprise에서는 admin/owner가 Remote Control을 workspace settings 또는 RBAC로 허용해야 할 수 있습니다. Windows host는 휴대폰이나 Mac에서 제어할 수 있지만, Windows Codex App 자체는 아직 다른 컴퓨터를 제어하는 클라이언트로 쓰지 못한다는 제한을 별도로 적었습니다.",
overview_update_4_title: "Enterprise 발표 맥락 분리",
overview_update_4_desc: "May 29 릴리스 노트에서 Windows Computer Use, Windows host remote control, responsiveness/browser stability, Codex Profiles를 확인했고, May 21 Enterprise/Edu 릴리스 노트와 May 22 Gartner 발표도 다시 대조했습니다. Goal mode GA, Appshots, locked computer use, in-app browser annotations, admin Codex analytics, plugin sharing 기본값, enterprise governance, sandboxing, RBAC, auditability를 일반 사용자 기능과 분리했습니다.",
overview_update_5_title: "관련 발표 전체 재점검",
overview_update_5_desc: "`developers.openai.com/codex`, `openai.com/index`, `openai.com/codex` 아래 관련 페이지를 다시 훑어봤습니다. 2026년 6월 2일 `Codex for every role, tool, and workflow` 발표는 role-specific plugins, Sites preview, 더 넓어진 annotations를 공식 발표 문구로 추가하고, 실제 활성화/운영 메모는 Sites 공식 문서와 pricing 문서 기준으로 분리했습니다. 2026년 5월 29일 Braintrust 고객 사례와 공식 `How OpenAI uses Codex` 자료는 고객 요청을 preview branch로 검증하는 흐름, Ask Mode, 이슈형 프롬프트, 환경 개선, AGENTS.md, Best-of-N 같은 실무 패턴으로 반영했습니다. 2026년 5월 27일 Tax AI/Codex engineering 글은 실무자 correction, production trace, eval target, bounded task environment, human review를 묶는 self-improving agent 운영 패턴으로 유지했습니다.",
overview_update_6_title: "Docs MCP·API changelog·설정 기준 재점검",
overview_update_6_desc: "Docs MCP, config reference, app/settings를 다시 확인해 project-scoped `.codex/config.toml`이 trusted project에서만 로드되고 machine-local provider/auth/profile/telemetry 키는 user-level config에 둬야 한다는 제한, shared CLI/IDE config, AGENTS.md 유도 문구, `review_model`, `plan_mode_reasoning_effort`, `approval_policy.granular`, `approvals_reviewer`, `default_permissions`, permissions profile, sandboxed network proxy/requirements, keyring credential storage, forced login/workspace, `sqlite_home`, hooks/plugins/skills 제어 키를 최신 흐름에 맞췄습니다. App `Settings > Profile`에는 lifetime tokens, peak tokens, streaks, longest task, token activity가 표시된다는 공식 설명도 유지했습니다. OpenAI API changelog와 deprecations 문서도 별도로 확인해 `gpt-5-codex`, `gpt-5.1-codex*`, `gpt-5.2-codex` API snapshot의 2026년 7월 23일 종료 예정과 `gpt-5.5` 대체 안내를 Platform/API 개발자 참고사항으로 분리했습니다. Secure MCP Tunnel, workload identity federation, Admin API 확장, multiple IP allowlists, Apps SDK의 MCP Apps lifecycle/server-instructions 변경은 Codex 일반 설정이 아니라 Platform/API/App developer lane에만 둡니다.",
overview_lens_beginner_kicker: "처음 사용자 한눈 요약",
overview_lens_beginner_title: "오늘 바로 실행할 최소 경로",
overview_lens_beginner_desc: "설치 이후 어디까지 해야 시작이 끝나는지 먼저 고정합니다.",
overview_lens_beginner_s1: "Quickstart 기준으로 앱/IDE/CLI 중 하나만 먼저 고릅니다.",
overview_lens_beginner_s2: "Local + 기본 sandbox/approval로 첫 작업 1건을 완료합니다.",
overview_lens_beginner_outcome: "결과: 첫 작업 성공 + 되돌릴 수 있는 Git 체크포인트",
overview_lens_pro_kicker: "실무 사용자 한눈 요약",
overview_lens_pro_title: "팀 운영 기준을 먼저 고정",
overview_lens_pro_desc: "개인 설정이 아니라 프로젝트 규칙으로 행동을 통일합니다.",
overview_lens_pro_s1: "AGENTS.md와 .codex/config.toml로 모델·권한·완료 기준을 명시합니다.",
overview_lens_pro_s2: "Docs MCP를 기본 연결해 제품/API 답변을 문서 근거 기반으로 유지합니다.",
overview_lens_pro_outcome: "결과: 팀별 재현 가능한 실행/리뷰/자동화 흐름",
showcase_kicker: "Built with Codex",
showcase_title: "Codex로 만든 실제 결과물을 먼저 둘러보세요",
showcase_desc: "OpenAI Developer Showcase에는 Codex로 만든 앱, 게임, 랜딩 페이지, 데이터 시각화 예제가 모여 있습니다. 어떤 결과물이 가능한지 먼저 둘러보고, 바로 실행해볼 수 있는 예시 프롬프트의 감도 함께 잡아보세요.",
showcase_cta: "Showcase 보기",
overview_decision_kicker: "Decision first",
overview_decision_title: "전공자가 바로 고를 수 있는 상황별 선택표",
overview_decision_desc: "긴 설명을 읽기 전에, 지금 하려는 작업에 맞는 사용 방식과 모델을 먼저 고를 수 있게 압축했습니다.",
overview_decision_th_case: "상황",
overview_decision_th_surface: "추천 사용 방식",
overview_decision_th_model: "모델 기준",
overview_decision_th_next: "바로 읽을 곳",
overview_decision_local_case: "내 프로젝트에서 작은 수정",
overview_decision_local_surface: "App Local 또는 CLI",
overview_decision_local_model: "gpt-5.5, 없으면 gpt-5.4",
overview_decision_local_next: "05-06 설치와 첫 실행",
overview_decision_ide_case: "IDE 안에서 계속 코딩",
overview_decision_ide_surface: "IDE Extension",
overview_decision_ide_model: "기본 추천 모델 + 프로젝트 컨텍스트",
overview_decision_ide_next: "08 IDE Extension",
overview_decision_team_case: "팀 규칙과 자동화 도입",
overview_decision_team_surface: "CLI/App + AGENTS.md + MCP",
overview_decision_team_model: "gpt-5.5/gpt-5.4, 빠른 보조 작업은 mini",
overview_decision_team_next: "12-16 운영 설정",
overview_decision_review_case: "PR 리뷰와 클라우드 작업",
overview_decision_review_surface: "Web/Cloud",
overview_decision_review_model: "Cloud/review 경로 기준 모델",
overview_decision_review_next: "09 Web + 16 자동화",
overview_beginner_tag: "처음 시작",
overview_beginner_title: "설치부터 첫 작업까지",
overview_beginner_desc: "Quickstart와 App 문서를 기준으로, 프로젝트 선택, Local 실행, 첫 요청, Git 체크포인트 순서로 익히는 경로입니다.",
overview_beginner_p1: "04-06에서 설치, 로그인, 첫 CLI/App 실행을 먼저 익힙니다.",
overview_beginner_p2: "10에서 sandbox와 approval이 왜 분리되는지 이해합니다.",
overview_beginner_p3: "14에서 OpenAI Docs MCP를 연결해 답변 신뢰도를 높입니다.",
overview_beginner_cta: "설치부터 보기",
overview_pro_tag: "실무 운영",
overview_pro_title: "팀 규칙과 자동화까지",
overview_pro_desc: "AGENTS.md, config.toml, 세션 운영, non-interactive 실행, 웹 리뷰 자동화를 연결해서 팀 작업 방식으로 확장하는 경로입니다.",
overview_pro_p1: "12-14에서 규칙, 설정, 연결할 도구와 권한을 프로젝트 단위로 정합니다.",
overview_pro_p2: "15-16에서 이어서 작업하기, codex exec, PR 리뷰 자동화를 한 흐름으로 묶습니다.",
overview_pro_p3: "17에서 출력 형식, 도구 사용, 검증 방식을 팀 표준 프롬프트로 정리합니다.",
overview_pro_cta: "운영 가이드 보기",
overview_visual_kicker: "Visual guide",
overview_visual_title: "그림으로 먼저 이해하는 Codex 시작 흐름",
overview_visual_desc: "처음 보는 사용자라면 이 그림만 먼저 훑어도 됩니다. 어디서 시작할지 고르고, 작은 작업으로 시작하고, sandbox와 Git으로 안전하게 진행한 다음, AGENTS.md와 Docs MCP로 점점 더 똑똑하게 쓰는 흐름을 한 장에 압축했습니다.",
overview_visual_meta: "이 안내 그림은 ImageGen 스킬로 제작했습니다.",
overview_visual_caption: "초보자용 요약 그림: 시작 방식 선택 → 작은 작업 → 안전하게 실행 → AGENTS.md와 Docs MCP로 확장",
overview_plain_kicker: "Plain language",
overview_plain_title: "전문 용어를 일단 이렇게 생각하면 쉽습니다",
overview_plain_1: "`Local`은 내 컴퓨터에서 일하는 것, `Cloud`는 원격 작업실에 맡기는 것이라고 보면 됩니다.",
overview_plain_2: "`Sandbox`는 작업 울타리, `Approval`은 울타리 밖으로 나가기 전에 한 번 더 묻는 버튼입니다.",
overview_plain_3: "`AGENTS.md`는 “이 프로젝트에서는 이렇게 일해 주세요”라고 적어두는 작업 설명서입니다.",
overview_plain_4: "`MCP`는 문서, 브라우저, 외부 도구를 Codex와 연결해 주는 케이블에 가깝습니다.",
overview_analogy_kicker: "Analogy",
overview_analogy_title: "Codex를 “옆자리 개발 보조”처럼 생각해도 됩니다",
overview_analogy_desc: "코드를 대신 전부 알아서 만드는 마법 상자라고 보기보다, 파일을 읽고 수정하고 테스트까지 도와주는 꽤 유능한 보조 개발자라고 생각하면 이해가 쉽습니다. 그래서 처음에는 작은 일부터 맡기고, 안전 장치와 프로젝트 규칙을 같이 붙이는 흐름이 가장 자연스럽습니다.",
overview_bridge_beginner_kicker: "처음 사용자 용어 가이드",
overview_bridge_beginner_title: "이렇게 읽으면 헷갈리지 않습니다",
overview_bridge_beginner_1: "Local은 내 컴퓨터에서 실행, Cloud는 원격 환경에서 실행입니다.",
overview_bridge_beginner_2: "Sandbox는 접근 범위, Approval은 실행 전 멈춤 기준입니다.",
overview_bridge_beginner_3: "MCP는 문서/도구를 연결해 답변 근거를 강화하는 레이어입니다.",
overview_bridge_pro_kicker: "실무 사용자 운영 포인트",
overview_bridge_pro_title: "팀 기준으로 보면 중요한 지점",
overview_bridge_pro_1: "프로젝트 신뢰 경계는 sandbox_mode와 writable_roots로 먼저 고정합니다.",
overview_bridge_pro_2: "작업 중 승인 개입 빈도는 approval_policy와 rules로 조정합니다.",
overview_bridge_pro_3: "질의 품질 일관성은 AGENTS.md와 Docs MCP 기본화로 확보합니다.",
overview_action_beginner_kicker: "처음 사용자 10분 플랜",
overview_action_beginner_title: "설치 후 바로 확인할 3가지",
overview_action_beginner_1: "codex 실행 후 현재 프로젝트 경로가 맞는지 먼저 확인합니다.",
overview_action_beginner_2: "sandbox/approval은 기본값으로 두고 첫 요청은 작은 수정부터 시작합니다.",
overview_action_beginner_3: "작업 전후 Git 체크포인트를 남겨 되돌리기 경로를 확보합니다.",
overview_action_beginner_cta: "초보 체크리스트 따라가기",
overview_action_pro_kicker: "실무 사용자 운영 플랜",
overview_action_pro_title: "팀에 붙일 때 먼저 고정할 3가지",
overview_action_pro_1: "AGENTS.md에 역할, 금지사항, 완료 조건을 명확히 선언합니다.",
overview_action_pro_2: "config.toml에서 모델, reasoning effort, 권한 정책을 팀 기본값으로 통일합니다.",
overview_action_pro_3: "OpenAI Docs MCP를 연결해 제품/API 질의는 문서 기반으로 답하도록 고정합니다.",
overview_action_pro_cta: "실무 체크리스트 따라가기",
overview_fact_kicker: "Officially verified today",
overview_fact_title: "오늘의 업데이트",
overview_fact_date_prefix: "업데이트 기준일",
overview_fact_1_title: "모델 추천과 실행 면 정렬",
overview_fact_1_desc: "`codex/models`를 오늘 다시 확인한 결과, 최우선 추천 모델은 여전히 `gpt-5.5`입니다. Codex 모델 선택기에 보이면 대부분의 작업을 거기서 시작하고, 아직 계정에 보이지 않으면 `gpt-5.4`를 대체 선택지로 사용합니다. 오늘 주의할 차이는 `codex/models`가 `gpt-5.5`의 API Access를 true로 표시하고 ChatGPT 또는 API-key authentication 경로를 언급하지만, `codex/pricing`의 API Key 사용량 표는 아직 `gpt-5.5`를 Not available로 둔다는 점입니다. 그래서 API-key 사용자는 실제 모델 선택기와 표준 API 가격표를 함께 확인해야 합니다. `gpt-5.4-mini`는 빠르고 효율적인 로컬/subagent 선택지이고, Codex Cloud는 여전히 `gpt-5.3-codex` 축에 남아 있으며, 대안 모델은 `gpt-5.2`입니다.",
overview_fact_2_title: "GPT-5.5 승격과 대체 선택지 구조",
overview_fact_2_desc: "오늘 기준 `Introducing GPT-5.5`, 최신 `codex/models`, API Models 문서를 함께 보면 `gpt-5.5`가 복잡한 코딩·computer use·knowledge work·research workflow의 최우선 추천 모델이라는 점이 유지됩니다. 다만 오늘 문서 정리에서는 “Codex 안에서 고르는 모델”, “API-key authentication으로 Codex를 쓰는 경우”, “직접 API에서 호출 가능한 모델”을 명확히 분리했습니다. API 자체에서는 `gpt-5.5`와 `gpt-5.5-pro`를 별도 모델로 사용할 수 있고, API deprecations 문서는 `gpt-5-codex` 계열 snapshot을 2026년 7월 23일 종료 예정으로 안내합니다. Codex API-key 사용자는 모델 페이지와 가격표의 차이를 함께 확인해야 합니다. `gpt-5.4`는 rollout 계정이나 비용/속도 기준에서 여전히 강력한 대체 선택지이고, `gpt-5.4-mini`는 빠른 작업과 subagent용 효율 모델로 남아 있습니다.",
overview_fact_3_title: "플랜/팀 요금제 보강",
overview_fact_3_desc: "Quickstart는 모든 ChatGPT 플랜에 Codex가 포함된다고 설명합니다. `codex/pricing`은 오늘도 Plus / Pro 5x / Pro 20x / Business / API Key 표와 Plus 로컬 메시지 범위(`gpt-5.5` 15-80, `gpt-5.4` 20-100, `gpt-5.4-mini` 60-350, `gpt-5.3-codex` 30-150)를 유지합니다. Plus는 web, CLI, IDE extension, iOS 접근을 함께 표시합니다. Sites는 현재 Business/Enterprise preview로 표시되고, pricing FAQ는 preview 기간 동안 무료라고 설명합니다. 다만 API-key 관점에서는 `codex/models`와 `codex/pricing`의 표현이 다르므로, 실무 문서는 사용량 표, token credit 표, usage dashboard, App `Settings > Profile`의 lifetime tokens/token activity, 실제 모델 선택기, API 모델 가격표를 섞지 말고 함께 확인하도록 고쳤습니다.",
overview_fact_4_title: "IDE·Windows 경로 재정리",
overview_fact_4_desc: "IDE 문서는 이제 VS Code 계열과 JetBrains 모두를 정식 경로로 다루고, macOS·Windows·Linux 사용을 함께 설명합니다. Windows 앱 문서는 Microsoft Store 또는 `winget install Codex -s msstore` 설치, PowerShell 기반 네이티브 agent, 네이티브 Windows sandbox, WSL2 전환 시 앱 재시작, 통합 터미널 선택을 명확히 나눕니다. 새 Windows sandbox 발표까지 함께 보면 `elevated`가 권장 native sandbox이고 `unelevated`는 관리자 승인이나 정책 때문에 elevated setup이 막힐 때의 fallback입니다.",
overview_fact_5_title: "App 운영 흐름 반영",
overview_fact_5_desc: "`Introducing the Codex app`와 2026년 4월 16일 `Codex for (almost) everything` 발표, 현재 `app/features`·`app/automations`·`app/settings`를 다시 대조해 worktrees, built-in Git, integrated terminal, skills, Automations, thread automations, chats, in-app browser, artifact preview, image generation뿐 아니라 background computer use, memories, context-aware suggestions, 더 넓어진 plugin 생태계, role-specific plugins, broader annotations, multiple terminal tabs, SSH devbox, summary pane 흐름까지 앞쪽에서 이해할 수 있게 재배치했습니다. Sites는 Business/Enterprise에서 OpenAI-hosted production URL을 저장/배포하는 preview로 추가했고, Enterprise RBAC/admin enablement, save version과 deploy version 구분, access mode, runtime secret은 Sites panel에서 관리해야 한다는 주의점을 함께 적었습니다. 2026년 6월 1일 OpenAI/AWS 발표와 Codex Amazon Bedrock 문서는 AWS-managed authentication, account controls, billing을 쓰는 enterprise/provider 배포 경로로 추가했고, ChatGPT 플랜 한도 변화로 쓰지 않았습니다.",
overview_fact_6_title: "config/API/plugin 흐름 갱신",
overview_fact_6_desc: "`review_model`, `approval_policy.granular`, `approvals_reviewer`, top-level `web_search`, `tools.web_search`, `tools.view_image`, `tool_suggest`, `service_tier`, `personality`, `default_permissions`, `windows.sandbox`, `windows.sandbox_private_desktop`, `model_instructions_file`, `memories.disable_on_external_context`, `sqlite_home`, TUI 설정, protected paths, JSON schema, app connector 권한, feature flag, permissions profile, sandboxed network proxy/requirements, keyring credential storage, forced login/workspace, OpenTelemetry export, plugin 배포 계층, Bedrock provider 설정, 로컬 OS 샌드박스와 cloud split runtime 설명을 함께 최신 기준으로 맞췄습니다. 특히 project-scoped config는 trusted project에서만 로드되고 provider/auth/profile/telemetry 라우팅 키는 user-level config에 둬야 합니다. OpenAI API changelog와 deprecations 문서는 Platform/API 개발자용 흐름으로 따로 읽어야 합니다. 예를 들어 OpenAI models on Amazon Bedrock는 OpenAI-compatible Responses API endpoint를 제공하지만 지원 모델과 기능은 AWS Region마다 다를 수 있고, `gpt-5-codex`, `gpt-5.1-codex`, `gpt-5.1-codex-max`, `gpt-5.2-codex` API snapshot은 2026년 7월 23일 종료 예정으로 표시되며 대체 모델은 `gpt-5.5`입니다. Secure MCP Tunnel은 customer-hosted outbound `tunnel-client`로 private MCP server를 연결하는 enterprise 옵션이며 initial GA는 account-led입니다.",
overview_sources_label: "근거 문서",
overview_source_1: "Codex Quickstart",
overview_source_16: "Codex Overview",
overview_source_2: "Codex Pricing",
overview_source_3: "Codex Best practices",
overview_source_4: "Codex Config Reference",
overview_source_5: "Agent approvals & security",
overview_source_6: "IDE extension",
overview_source_7: "Windows guide",
overview_source_20: "Docs MCP",
overview_source_8: "Codex Models",
overview_source_9: "GPT-5.5 announcement",
overview_source_14: "GPT-5.4 mini and nano announcement",
overview_source_10: "GPT-5.3-Codex-Spark announcement",
overview_source_13: "Codex App announcement",
overview_source_17: "Codex App Automations",
overview_source_18: "Codex App Features",
overview_source_19: "Codex Windows App",
overview_source_15: "Codex Plugins",
overview_source_21: "Codex Skills",
overview_source_11: "Codex for OSS",
overview_source_12: "AI-Native Engineering Team",
overview_source_22: "Codex for (almost) everything",
overview_source_23: "OpenAI API Models",
overview_source_24: "GPT-5.5 pro API Model",
overview_source_25: "Codex In-app Browser",
overview_source_26: "Pay-as-you-go pricing for teams",
overview_source_27: "Codex rate card",
overview_source_28: "Codex on Amazon Bedrock",
overview_source_29: "Advanced Account Security",
overview_source_30: "GPT-5.5 Instant",
overview_source_31: "Running Codex safely",
overview_source_32: "GPT-5.5-Cyber and Codex Security",
overview_source_33: "OpenAI API Changelog",
overview_source_34: "Codex Sites",
s1_refs: '출처: <a href="https://developers.openai.com/codex/" target="_blank">Codex overview</a> · <a href="https://developers.openai.com/codex/models/" target="_blank">Codex models</a> · <a href="https://developers.openai.com/api/docs/models" target="_blank">OpenAI API Models</a> · <a href="https://developers.openai.com/api/docs/changelog" target="_blank">OpenAI API changelog</a> · <a href="https://openai.com/index/introducing-gpt-5-5/" target="_blank">Introducing GPT-5.5</a> · <a href="https://openai.com/index/work-with-codex-from-anywhere/" target="_blank">Work with Codex from anywhere</a>',
s2_refs: '출처: <a href="https://developers.openai.com/codex/quickstart/" target="_blank">Quickstart</a> · <a href="https://developers.openai.com/codex/app" target="_blank">Codex app</a> · <a href="https://developers.openai.com/codex/ide" target="_blank">IDE extension</a> · <a href="https://developers.openai.com/codex/cli" target="_blank">Codex CLI</a>',
s3_refs: '출처: <a href="https://developers.openai.com/codex/models/" target="_blank">Codex models</a> · <a href="https://developers.openai.com/codex/pricing/" target="_blank">Codex pricing</a> · <a href="https://developers.openai.com/api/docs/models" target="_blank">OpenAI API Models</a> · <a href="https://developers.openai.com/api/docs/models/gpt-5.5-pro" target="_blank">GPT-5.5 pro API model</a>',
s4_refs: '출처: <a href="https://developers.openai.com/codex/quickstart/" target="_blank">Quickstart</a> · <a href="https://developers.openai.com/codex/pricing/" target="_blank">Codex pricing</a> · <a href="https://help.openai.com/en/articles/20001106-codex-rate-card" target="_blank">Codex rate card</a> · <a href="https://developers.openai.com/codex/speed" target="_blank">Speed</a> · <a href="https://openai.com/index/codex-flexible-pricing-for-teams/" target="_blank">Pay-as-you-go pricing for teams</a>',
s5_refs: '출처: <a href="https://developers.openai.com/codex/quickstart/" target="_blank">Quickstart</a> · <a href="https://developers.openai.com/codex/windows" target="_blank">Windows guide</a> · <a href="https://developers.openai.com/codex/app/windows" target="_blank">Windows app</a> · <a href="https://developers.openai.com/codex/ide" target="_blank">IDE extension</a> · <a href="https://openai.com/index/building-codex-windows-sandbox/" target="_blank">Windows sandbox engineering</a> · <a href="https://openai.com/index/running-codex-safely/" target="_blank">Running Codex safely</a>',
s6_refs: '출처: <a href="https://developers.openai.com/codex/quickstart/" target="_blank">Quickstart</a> · <a href="https://developers.openai.com/codex/cli" target="_blank">Codex CLI</a> · <a href="https://developers.openai.com/codex/cli/slash-commands/" target="_blank">CLI slash commands</a>',
s7_refs: '출처: <a href="https://developers.openai.com/codex/app" target="_blank">Codex app</a> · <a href="https://developers.openai.com/codex/app/features" target="_blank">App features</a> · <a href="https://developers.openai.com/codex/app/browser" target="_blank">In-app browser</a> · <a href="https://developers.openai.com/codex/app/automations" target="_blank">App automations</a> · <a href="https://openai.com/index/codex-for-almost-everything/" target="_blank">Codex for (almost) everything</a>',
s8_refs: '출처: <a href="https://developers.openai.com/codex/ide" target="_blank">IDE extension</a> · <a href="https://developers.openai.com/codex/ide/features" target="_blank">IDE features</a> · <a href="https://developers.openai.com/codex/ide/slash-commands" target="_blank">IDE slash commands</a> · <a href="https://developers.openai.com/codex/app/windows" target="_blank">Windows app</a>',
s9_refs: '출처: <a href="https://developers.openai.com/codex/quickstart/" target="_blank">Quickstart</a> · <a href="https://developers.openai.com/codex/cloud" target="_blank">Codex cloud</a> · <a href="https://developers.openai.com/codex/remote-connections" target="_blank">Remote connections</a> · <a href="https://developers.openai.com/codex/integrations/github" target="_blank">GitHub integration</a> · <a href="https://openai.com/index/work-with-codex-from-anywhere/" target="_blank">Work with Codex from anywhere</a> · <a href="https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes" target="_blank">Enterprise/Edu release notes</a>',
s10_refs: '출처: <a href="https://developers.openai.com/codex/agent-approvals-security" target="_blank">Agent approvals & security</a> · <a href="https://developers.openai.com/codex/config-reference/" target="_blank">Config reference</a> · <a href="https://developers.openai.com/codex/windows" target="_blank">Windows guide</a> · <a href="https://openai.com/index/building-codex-windows-sandbox/" target="_blank">Windows sandbox engineering</a>',
changelog_kicker: "Changelog",
changelog_title: "오늘의 업데이트",
changelog_item_1: "`gpt-5.5` 우선 추천을 다시 확인하고, `codex/models`의 API Access 표기와 `codex/pricing`의 API Key 사용량 표 차이를 분리했습니다.",
changelog_item_2: "가격 설명에서 Pro $100 2x 만료일과 Pro $200의 20x/25x Plus 한도 구분, Go/iOS 표기, credit 적용 범위, API-key 사용 시 실제 모델 선택기와 직접 API 모델 가격 확인 주의점을 맞췄습니다.",
changelog_item_3: "Remote connections 공식 문서가 Windows host를 ChatGPT iOS/Android 또는 Mac에서 제어할 수 있다고 바뀐 점을 반영하고, Windows App은 아직 다른 컴퓨터를 제어하는 클라이언트가 아니라는 제한을 분리했습니다.",
changelog_item_4: "May 29 릴리스 노트 기준 Windows Computer Use, Windows host remote control, Codex Profiles, responsiveness/browser stability를 반영했습니다.",
changelog_item_5: "2026년 6월 2일 role-specific plugins, Sites preview, broader annotations 발표를 추가하고, Bedrock과 Platform/API 항목은 별도 lane에 유지했습니다.",
changelog_item_6: "Chrome 소셜 확인에서 X/LinkedIn의 공식 발표 반복은 확인했고, Threads에서는 공식 문서로 반영할 새 claim을 찾지 못했습니다. 검증되지 않은 소셜 claim은 추가하지 않았습니다.",
overview_update_7_title: "소셜 팁 검증 제한 기록",
changelog_sources_toggle: "출처 배치 방식 보기",
changelog_sources_note: "이번부터는 긴 출처 묶음을 한곳에 모으는 대신, 각 섹션 아래의 `출처` 줄에서 바로 확인할 수 있게 정리했습니다. 위 changelog는 변경 요약만 읽고, 세부 근거는 관련 섹션에서 바로 여는 흐름을 권장합니다.",
// Section 1
s1_title: "Codex 한눈에 보기",
s1_intro: "OpenAI Codex는 코드를 읽고, 고치고, 테스트하고, 결과를 설명하는 코딩 에이전트입니다.",
s1_philosophy: '"코드를 쓰는 도구 → 작업을 끝내는 에이전트"',
s1_desc: "모델, 요금, 설정은 뒤에서 자세히 보고, 여기서는 Codex가 어떤 표면으로 확장되어 왔는지만 빠르게 잡습니다.",
s1_timeline_title: "핵심 업데이트",
th_date: "시기",
th_event: "이벤트",
tl1: "Codex CLI 오픈소스 공개",
tl2: "클라우드 샌드박스와 Codex-1 공개",
tl3: "Codex CLI Rust 리라이트 베타 공개",
tl4: "DevDay 2025에서 GPT-5 Codex GA 발표",
tl5: "GPT-5.2-Codex 출시 (멀티파일 최적화, 컨텍스트 압축)",
tl6: "IDE Extension으로 편집기 안 사용 흐름 확장",
tl7: "Codex macOS App 공개",
tl8: "GPT-5.3-Codex 출시",
tl9: "Codex-Spark 리서치 프리뷰 공개",
tl10: "Windows App으로 데스크톱 표면 확장",
tl11: "GPT-5.4 공개 및 Codex 추천 기본 모델로 반영",
tl12: "GPT-5.5가 Codex 최우선 추천 모델로 승격",
tl13: "ChatGPT 모바일 앱의 Codex 원격 조율 preview 공개",
// Section 2
s2_title: "Codex 제품군 한눈에 보기",
s2_cli: "터미널 기반(TUI) 로컬 에이전틱 코딩 및 CI/CD 자동화. 2025.04 출시.",
s2_app: "데스크톱 앱(macOS + Windows). 멀티에이전트 커맨드 센터, 내장 worktree, integrated terminal, diff review, skills/automations 관리, personality 전환까지 한곳에서 다룹니다.",
s2_ide: "VS Code, Cursor, Windsurf, JetBrains로 확장된 IDE 에이전트. 로컬 작업과 클라우드 위임을 함께 지원합니다.",
s2_cloud: "ChatGPT 웹/모바일 표면. 웹에서는 GitHub 저장소 환경 연결, 클라우드 작업 실행, diff 리뷰, 자동 PR 코드 리뷰, PR 댓글의 `@codex` 위임을 담당하고, 모바일 preview에서는 Codex가 실행 중인 Mac/원격 환경의 thread, 승인, model 변경, screenshot/terminal/diff/test output 확인을 이어갑니다.",
s3_title: "지원 모델",
s3_models_title: "현재 모델 가이드",
s3_rec_title: "추천 모델",
s3_alt_title: "대안 모델",
th_model: "모델", th_feature: "특징", th_use: "용도", th_note: "비고",
m0_f: "복잡한 코딩, computer use, 지식 작업, 리서치 워크플로를 위한 최신 프론티어 모델. Codex 모델 선택기에서 보이면 최우선 추천 모델입니다.", m0_u: "가장 중요한 고난도 작업",
m1_f: "GPT-5.5가 아직 모델 선택기에 보이지 않거나 비용/속도 균형이 더 중요할 때의 기본 대체 모델입니다. 강한 코딩·추론·도구 사용 균형이 계속 장점이고, built-in computer use가 처음 본격적으로 강조된 mainline 모델이기도 합니다.", m1_u: "전문 개발 작업과 폭넓은 워크플로",
m1mini_f: "빠르고 효율적인 mini 모델로, 가벼운 코딩 작업과 subagent/보조 agent 작업에 적합합니다.",
m2_f: "복잡한 소프트웨어 엔지니어링에 강한 업계 최고 수준의 코딩 모델입니다. 이 모델의 코딩 역량은 현재 GPT-5.4에도 반영되어 있습니다.", m2_u: "복잡한 실제 SW 엔지니어링",
m3_f: "텍스트 전용 리서치 프리뷰 모델. 거의 즉각적인 실시간 코딩 반복에 최적화되었으며 ChatGPT Pro 사용자에게 제공됩니다.", m3_u: "실시간 코딩, 빠른 반복",
s3_rec_note: "💡 최신 `codex/models` 기준 공식 가이드는 모델 선택기에서 `gpt-5.5`가 보이면 대부분의 작업을 거기서 시작하라고 권장합니다. 오늘은 같은 모델 페이지가 API Access를 true로 표시하지만 `codex/pricing`의 API Key 사용량 표는 아직 Not available로 두고 있으므로, API-key 사용자는 실제 Codex 모델 선택기와 표준 API pricing을 같이 확인하세요. 비용/속도 균형이 중요하면 `gpt-5.4`를 계속 사용하면 됩니다. API에서는 `gpt-5.5`와 `gpt-5.5-pro`를 사용할 수 있으며, Pro는 더 높은 정확도가 필요한 Responses API/Batch API 작업에 적합합니다. 더 빠르고 가벼운 코딩 작업이나 subagent에는 `gpt-5.4-mini`, 복잡한 소프트웨어 엔지니어링과 현재 Codex Cloud에는 `gpt-5.3-codex`, 거의 즉각적인 반복에는 `gpt-5.3-codex-spark`를 검토하세요. 현재 대표 대안 모델은 `gpt-5.2`입니다.",
computer_use_kicker: "Computer use",
computer_use_title: "Computer use를 하나의 기능 챕터로 이해하기",
computer_use_desc: "공식 문서 기준으로 `gpt-5.4`는 첫 mainline built-in computer use 모델입니다. 즉, 화면을 보고 다음 행동을 정하고, 클릭이나 입력을 실행하고, 바뀐 화면을 다시 확인하는 루프를 모델 자체가 더 잘 돌 수 있게 됐다는 뜻입니다.",
computer_use_p1: "단순 채팅형 답변보다 “직접 UI를 보면서 작업”하는 자동화에 더 잘 맞습니다.",
computer_use_p2: "핵심 흐름은 `screenshot → action → screenshot → verify/fix` 입니다.",
computer_use_p3: "브라우저 테스트, 폼 입력, 단계가 많은 업무 플로우 검증에서 특히 설명이 쉬워집니다.",
computer_use_p4: "`Plugins`가 외부 서비스를 연결하는 도구 상자라면, `computer use`는 눈과 손으로 화면을 직접 다루는 기능에 가깝습니다.",
computer_use_p5: "`Skills`는 일을 하는 방법을 설명하는 설명서이고, `computer use`는 그 설명서를 따라 실제 버튼을 누르고 입력하는 실행 수단입니다.",
computer_use_cta: "공식 Computer use 가이드 보기",
computer_use_caption: "영어는 OpenAI 공식 다이어그램, 한국어는 이해를 돕기 위한 설명 그림입니다.",
m4_f: "코딩과 에이전틱 작업 전반을 위한 이전 범용 모델. 현재는 `gpt-5.5`/`gpt-5.4` 조합 아래의 대안 모델로 보는 편이 맞습니다.", m4_u: "산업/도메인 전반",
s3_other_note: "⚠️ Codex는 위 모델들과 가장 잘 작동합니다. Chat Completions 또는 Responses API를 지원하는 다른 모델도 사용 가능하지만, Chat Completions API 지원은 향후 제거 예정입니다.",
model_show: "상세 보기", model_hide: "접기",
feat_cap: "성능", feat_spd: "속도",
feat_cli: "Codex CLI & SDK", feat_app: "Codex 앱 & IDE 확장",
feat_cloud: "Codex Cloud", feat_credits: "ChatGPT 크레딧", feat_api: "API 접근",
s3_cfg_title: "🛠️ 모델 설정 팁",
s3_cfg_default: "config.toml에서 기본 모델을 설정할 수 있습니다. 지정하지 않으면 추천 모델이 기본값입니다.",
s3_cfg_temp: "CLI에서 <code>/model</code> 명령 또는 <code>codex -m 모델명</code> 플래그로 임시 변경 가능. IDE에서는 입력창 아래 모델 선택기 사용.",
s3_cfg_cloud: "Codex 클라우드 작업의 기본 모델은 현재 변경할 수 없습니다.",
// Section 3
s4_title: "시스템 요구사항 & 가격",
s4_req_title: "시스템 요구사항", s4_price_title: "주요 플랜과 접근 방식",
s4_git: "강력히 권장", s4_acct_label: "계정", s4_custom: "맞춤 가격",
s4_note_access: "현재 공식 문서는 모든 ChatGPT 플랜에 Codex가 포함된다고 설명합니다. 처음 보는 사용자라면 `Free $0`, `Go $8`, `Plus $20`, `Pro from $100` 순으로 보면 이해가 쉽고, 팀 사용자는 Business·Enterprise의 Codex-only pay-as-you-go seat를 같이 보면 됩니다. Pricing 페이지는 Plus / Pro 5x / Pro 20x / Business / API Key 표와 모델별 5시간 한도를 보여주며, Plus 기준 로컬 메시지 범위는 `gpt-5.5` 15-80, `gpt-5.4` 20-100, `gpt-5.4-mini` 60-350, `gpt-5.3-codex` 30-150입니다. 중요한 차이는 오늘 `codex/models`가 `gpt-5.5` API Access를 표시하지만 `codex/pricing`의 API Key 사용량 표는 아직 Not available로 남아 있다는 점입니다. 실제 비용 판단은 메시지/작업 한도, token credit rate card, usage dashboard, 실제 모델 선택기, API 모델 가격표를 섞지 말고 함께 대조하세요. 이미지 생성은 일반 Codex 한도를 평균 3-5배 더 빨리 쓰며 Fast mode는 지원 모델에서 더 높은 비율로 크레딧을 소모합니다.",
th_plan: "플랜", th_price: "가격", th_codex: "Codex 범위",
s4_codex_free: "빠른 코딩 작업 위주의 Codex 탐색",
s4_codex_go: "가벼운 코딩 작업을 위한 저가형 Codex 사용",
s4_codex_plus: "웹/CLI/IDE/iOS + 최신 모델(`gpt-5.5` 포함) + 클라우드 리뷰/Slack + 토큰 기반 크레딧 확장",
s4_codex_pro: "Spark 리서치 프리뷰 + priority processing + Pro 5x / Pro 20x 표 + 2026년 5월 31일까지였던 부스트는 dashboard 확인",
s4_codex_business: "표준 Business seat 또는 Codex-only pay-as-you-go seat + 더 큰 VM + credits + SAML SSO/MFA",
s4_codex_enterprise: "Business 포함 + Codex-only pay-as-you-go 가능 + 우선 처리 + RBAC/감사/거버넌스",
s4_usage_based: "토큰 사용량 기반",
s4_codex_api: "CLI/IDE/SDK 전용, 클라우드 기능 없음. API-key 사용 시 실제 모델 선택기와 표준 API 가격표를 함께 확인",
// Section 4
s5_title: "설치 및 인증",
s5_cli_install: "CLI 설치", s5_auth: "인증",
s5_app_install: "Desktop App 설치 (macOS/Windows)",
s5_app1: "macOS: openai.com/codex 에서 다운로드 (.dmg). Intel Mac 사용자는 Intel 빌드를 선택합니다.",
s5_app2: "Windows: Microsoft Store에서 설치하거나 `winget install Codex -s msstore` 사용",
s5_app3: "앱 실행 후 ChatGPT 계정 또는 API 키로 로그인 (API 키 로그인은 cloud threads와 일부 credits 기반 기능이 제한될 수 있음). 고위험 계정이나 민감한 프로젝트라면 ChatGPT 웹 보안 설정에서 Advanced Account Security도 검토하세요.",
s5_waitlist_tag: "Windows App",
s5_waitlist_scope: "지원 범위는 수시 변경",
s5_waitlist_text: "업데이트: Windows에서는 Codex 앱이 가장 쉬운 기본 경로입니다. Microsoft Store나 `winget install Codex -s msstore`로 설치할 수 있고, 기본 agent는 PowerShell에서 네이티브로 실행됩니다. Default permissions를 켜면 Windows sandbox가 적용되고, Linux 툴체인이나 WSL 파일시스템 프로젝트가 중요하면 Settings에서 agent를 WSL로 바꾼 뒤 앱을 재시작하세요. 통합 터미널은 PowerShell, Command Prompt, Git Bash, WSL 중 별도로 고를 수 있습니다.",
s5_waitlist_btn: "👉 Windows App 공식 문서 보기",
s5_vsc_install: "IDE Extension 설치",
s5_vsc1: 'VS Code/Cursor/Windsurf는 확장을 설치하고, JetBrains는 전용 통합을 사용합니다. JetBrains 쪽은 ChatGPT, API 키, JetBrains AI 로그인을 지원하고, Windows에서는 네이티브 sandbox 또는 WSL2 경로를 선택할 수 있습니다.',
s5_vsc2: "설치 후 사이드바에서 Codex 패널 활성화",
s5_vsc3: "ChatGPT 계정 또는 API 키로 인증",
// Section 5
s6_title: "Codex CLI 사용법",
img_cli: "Codex CLI — 터미널 기반 에이전틱 코딩 워크플로 (2026년 4월 23일 촬영)",
// Section 6
s7_title: "Codex App (macOS + Windows)",
s7_intro: 'Codex App은 여러 프로젝트를 넘나들며 병렬 agent thread를 운영하고, built-in Git와 integrated terminal로 결과를 리뷰하며, chats·artifact preview·in-app browser·browser use·computer use·image generation·memories·thread automations를 한곳에서 다루는 데 가장 적합한 데스크톱 표면입니다. 최신 App 문서와 2026년 4월 업데이트 기준으로 multiple terminal tabs, richer summary/sidebar, 더 넓어진 plugins, SSH devbox, IDE sync, projectless chats 흐름까지 함께 보는 것이 현재 상태에 가깝고, 로컬 작업은 기본적으로 cached web search를 씁니다. Settings > Profile에서는 lifetime tokens, peak tokens, streaks, longest task, token activity 같은 사용 통계도 확인할 수 있으며, Windows에서는 네이티브 sandbox 또는 WSL2 경로를 선택할 수 있습니다.',
s7_modes_title: "쓰레드 모드 (Thread Modes)",
s7_th_where: "실행 위치", s7_th_desc: "설명",
s7_mode_local_where: "내 맥", s7_mode_local: "현재 프로젝트 디렉토리에서 직접 작업",
s7_mode_wt_where: "내 맥", s7_mode_wt: "Git Worktree로 변경사항 격리. 병렬 작업에 적합",
s7_mode_cloud_where: "클라우드", s7_mode_cloud: "원격 클라우드 환경에서 실행",
s7_feat_title: "핵심 기능",
s7_f1: "프로젝트 간 병렬 쓰레드 실행", s7_f2: "내장 Git 도구 (diff, commit, PR)",
s7_f_skills: "Skills & Plugins 지원", s7_f_auto: "Automations (정기 작업 자동화)",
s7_f_voice: "Voice Dictation (Ctrl+M)", s7_f_float: "플로팅 팝아웃 창",
s7_f_sync: "IDE Extension 실시간 동기화", s7_f_img: "이미지 입력 (Drag & Drop)",
s7_f_browser: "In-app browser, Browser use, 브라우저 코멘트", s7_f_computer: "Computer use (GUI 조작)",
s7_f_imagegen: "이미지 생성/편집", s7_f_memory: "Memories & thread automations",
s7_f_term: "내장 터미널 (쓰레드별)", s7_f_sleep: "작업 중 절전 방지",
s7_playbook_title: "이번 대규모 업데이트 기능 실전 가이드",
s7_playbook_intro: "아래 블록은 단순 소개가 아니라 “언제 쓰고, 어떻게 시작하고, 무엇을 주의해야 하는지”를 기능별로 압축한 실전 챕터입니다. 이번 릴리스에서 중요도가 큰 기능부터 읽으면 됩니다.",
browser_use_kicker: "Browser use",
browser_use_title: "Codex가 인앱 브라우저를 직접 조작하게 하기",
browser_use_desc: "Browser use는 In-app browser 위에서 Codex가 직접 클릭, 입력, 렌더 상태 검사, 스크린샷, 수정 검증까지 수행하게 하는 기능입니다. 로컬 개발 서버나 파일 기반 프리뷰처럼 로그인 없이 열 수 있는 페이지에서 프론트엔드 버그를 재현하고 고칠 때 가장 실전적입니다.",
browser_use_p1: "In-app browser는 함께 보는 화면이고, Browser use는 Codex가 그 화면을 직접 다루는 실행 모드입니다.",
browser_use_p2: "Browser 플러그인을 설치·활성화한 뒤 작업에서 `@Browser`를 명시하거나 “브라우저를 사용해서 확인해줘”라고 요청합니다.",
browser_use_p3: "로그인, 쿠키, 확장 프로그램, 기존 브라우저 프로필이 필요한 페이지에는 맞지 않습니다. 인증 없는 localhost, file preview, 공개 페이지에 쓰세요.",
browser_use_p4: "요청에는 URL, 관심 상태(loading/empty/error/success), 고칠 요소, 검증할 화면 크기를 같이 적는 것이 좋습니다.",
browser_use_cta: "공식 Browser use 문서 보기",
browser_use_recipe_kicker: "Example prompt",
browser_use_example: "<code>Use @Browser to open http://localhost:3000/settings.\nReproduce the mobile overflow bug at 390px width, fix only the overflowing controls,\nthen verify the page again and attach the screenshot result.</code>",
s7_pb_browser_t: "In-app browser",
s7_pb_browser_d: "로컬 개발 서버나 공개 페이지를 렌더 상태 그대로 검토할 때 씁니다. 로그인 없는 URL만 열고, 브라우저 코멘트로 특정 요소를 찍은 뒤 “이 코멘트만 반영해 달라”고 좁게 요청하는 흐름이 가장 안정적입니다.",
s7_pb_browser_use_t: "Browser use",
s7_pb_browser_use_d: "Codex가 직접 브라우저를 열고 클릭·입력·검증해야 할 때 씁니다. Browser 플러그인과 사이트 허용 설정이 필요하고, 작업은 한 URL·한 상태·한 버그처럼 작게 잘라야 결과를 검토하기 쉽습니다.",
s7_pb_computer_t: "Computer use",
s7_pb_computer_d: "CLI나 plugin으로 다루기 어려운 GUI 작업에 씁니다. macOS에서는 plugin 설치 후 Screen Recording과 Accessibility 권한을 주고, Windows에서는 2026년 5월 29일 릴리스 노트 기준 eligible user가 Computer Use를 쓸 수 있습니다. 단, Windows Computer Use는 출시 시점에 EEA, UK, Switzerland에서 제공되지 않으므로 지역/플랜 제한을 확인하고, 한 번에 한 앱·한 플로우만 맡기는 식으로 범위를 좁혀야 안전합니다.",
s7_pb_image_t: "Image generation",
s7_pb_image_d: "배너, 배경, UI 목업 자산, 일러스트를 코드 작업과 같은 thread에서 바로 만들거나 수정할 때 적합합니다. 작은 단위 자산은 내장 기능으로, 대량 생성이나 비용 분리는 API 키 경로로 넘기는 식으로 구분하면 좋습니다.",
s7_pb_memory_t: "Memories",
s7_pb_memory_d: "반복해서 설명하는 선호, 스택, 프로젝트 관례를 Codex가 다시 기억하게 할 때 유용합니다. 다만 팀의 필수 규칙은 여전히 `AGENTS.md`에 두고, memory는 개인 리콜 레이어로 쓰는 것이 공식 권장 흐름입니다.",
s7_pb_thread_t: "Thread automations",
s7_pb_thread_d: "같은 대화 문맥을 유지한 채 주기적으로 다시 깨워야 하는 작업에 씁니다. 긴 실행 상태 확인, PR 댓글 재점검, Slack/GitHub 폴링처럼 “이전 대화의 맥락”이 중요한 경우에 project automation보다 thread automation이 맞습니다. 현재 문서 기준으로 thread automation은 분 단위 간격이나 일/주간 스케줄에 잘 맞습니다.",
s7_pb_plugin_t: "Plugins",
s7_pb_plugin_d: "skills, app integrations, MCP server를 묶어 배포하는 단위입니다. 반복 프롬프트가 안정화된 뒤 팀에 공유할 때 plugin으로 올리고, 설치 후에는 작업 설명만 하거나 `@plugin`으로 특정 workflow를 직접 호출하는 방식이 좋습니다.",
img_app: "Codex Desktop App — 멀티에이전트 커맨드 센터와 워크트리 관리 (2026년 4월 23일 촬영)",
// Section 7
s8_title: "Codex IDE Extension",
s8_intro: "VS Code, Cursor, Windsurf, JetBrains IDE에 통합되는 AI 코딩 에이전트입니다. 일반적으로 설치 후 Agent mode가 기본이며, JetBrains 통합은 ChatGPT 계정, API 키, JetBrains AI로 로그인할 수 있습니다. 현재 공식 문서는 IDE 통합이 macOS·Windows·Linux에 걸쳐 제공된다고 설명합니다. Windows에서는 네이티브 sandbox를 기본 경로로 두고, Linux-native 도구나 이미 WSL2에 있는 저장소가 중요할 때 WSL2 워크스페이스를 선택하세요.",
th_mode: "모드", th_file_access: "파일 접근", th_edit: "코드 수정", th_exec: "명령 실행",
s8_chat_file: "△ 수동 컨텍스트만 (@file/열린 파일/선택 영역)",
s8_approval: "✅ (승인 필요)", s8_approval2: "✅ (승인 필요)",
s8_mode_note: "Chat 모드는 자동으로 파일을 탐색/수정하지 않지만, @file·열린 파일·선택 영역을 컨텍스트로 포함해 질문할 수 있습니다.",
s8_key_title: "핵심 기능",
s8_f_context: "@file 에디터 컨텍스트 자동 참조", s8_f_cloud: "클라우드 위임 (Cloud Delegation)",
s8_f_effort: "추론 노력 조절 (low/medium/high/xhigh)", s8_f_jetbrains: "JetBrains IDE 지원",
s8_f_cursor: "Cursor, Windsurf 호환", s8_f_sync: "App 실시간 동기화 (Auto Context)",
img_ide: "Codex IDE Extension — 마켓플레이스 (4.6M+ 다운로드)",
// Section 8
s9_title: "Codex Web/Mobile (ChatGPT)",
s9_intro: "ChatGPT 웹의 `chatgpt.com/codex`에서는 GitHub 환경을 연결하고 클라우드 코딩 작업을 실행할 수 있습니다. Remote connections는 별도 축입니다. 현재 공식 문서는 ChatGPT iOS/Android 모바일 앱이 awake/online macOS 또는 Windows Codex App host를 제어할 수 있다고 설명합니다. 모바일에서는 active thread, 승인, model 변경, 새 작업 시작, screenshot/terminal output/diff/test result 확인을 이어갈 수 있지만, 실제 파일·credential·permission·plugin·browser setup·Computer Use·local tool은 연결된 host에 남습니다. 같은 account/workspace와 필요한 SSO/MFA/passkey가 맞아야 하고, Business/Enterprise에서는 admin/owner가 Remote Control을 workspace settings 또는 RBAC로 허용해야 할 수 있습니다. Mac Codex App도 Windows host를 제어할 수 있지만, Windows Codex App 자체는 아직 다른 컴퓨터를 제어하는 클라이언트가 아니라는 제한을 별도로 기억하세요.",
s9_s1: "GitHub 계정 연결 → 레포 선택", s9_s2: "작업 설명 입력",
s9_s3: "클라우드 샌드박스에서 자동 작업", s9_s4: "결과 diff 확인", s9_s5: "GitHub PR로 제안 또는 머지",
s9_auto_title: "자동 코드 리뷰",
s9_auto_desc: 'Codex Web에서 GitHub 연결 후 리뷰 대상 저장소를 선택하고 "Automatic reviews"를 활성화하면 새 PR마다 자동 리뷰됩니다. AGENTS.md 규칙 기반 커스터마이징 가능.',
img_cloud: "Codex Web — ChatGPT 웹 대시보드에서 GitHub 연동 작업",
// Section 9
s10_title: "승인 + 샌드박스 조합",
th_read: "파일 읽기", th_write: "파일 수정", th_net: "네트워크",
default_badge: "기본",
s10_sandbox: "보안: 로컬 Codex는 OS 수준 sandbox를 사용합니다. macOS는 Seatbelt, Linux는 bubblewrap + seccomp, Windows는 네이티브 sandbox 또는 WSL을 사용합니다. Windows native에서는 `elevated`가 권장 경로이고, 관리자 승인이나 정책 때문에 막히면 `unelevated` fallback을 씁니다. 기본 `workspace-write`에서는 네트워크가 꺼져 있고 `.git`·`.codex`·`.agents`는 보호 경로로 읽기 전용입니다. Cloud는 setup 단계와 offline agent 단계를 분리한 런타임입니다.",
// Section 10
s11_title: "슬래시 명령어 & 단축키",
s11_cmd_title: "플랫폼별 슬래시 명령어", s11_key_title: "키보드 단축키 (CLI)",
s11_scope_note: "아래 목록은 CLI / IDE / App 공식 문서를 교차 확인해 분리한 명령어입니다. 서로 이름이 비슷해도 동작이 다를 수 있습니다.",
s11_key_scope: "아래 단축키는 CLI(TUI) 기준입니다. App은 OS별 단축키(macOS Cmd, Windows Ctrl)를, IDE는 각 IDE 키맵을 사용합니다.",
s11_cli_title: "CLI (터미널)",
s11_cli_c1: "활성 모델 변경 (가능한 경우 reasoning effort 포함)",
s11_cli_c2: "승인/권한 정책 변경",
s11_cli_c3: "활성 에이전트 스레드 전환",
s11_cli_c4: "플랜 모드 전환 (선택적으로 인라인 프롬프트 포함)",
s11_cli_c5: "대화 요약으로 컨텍스트 토큰 절약",
s11_cli_c6: "세션 설정/토큰 사용량 확인",
s11_cli_c7: "워크트리 코드 리뷰 실행",
s11_cli_c8: "진단 로그와 피드백 전송",
s11_cli_note: "CLI에는 이외에도 /plugins, /personality, /experimental, /init, /mention, /mcp, /diff, /fork, /resume, /new, /quit 등이 있습니다. /approvals는 별칭으로 동작하지만 팝업 목록에는 표시되지 않습니다.",
s11_ide_title: "IDE Extension",
s11_ide_c1: "자동 컨텍스트 포함 토글",
s11_ide_c2: "클라우드 모드로 전환",
s11_ide_c3: "클라우드 환경 선택 (cloud 모드에서만)",
s11_ide_c4: "로컬 모드로 전환",
s11_ide_c5: "변경사항 리뷰 모드 시작",
s11_ide_c6: "스레드 ID/컨텍스트/레이트리밋 확인",
s11_ide_c7: "피드백 다이얼로그 열기",
s11_app_title: "Codex App (Desktop)",
s11_app_c1: "멀티스텝 계획용 plan mode 토글",
s11_app_c2: "연결된 MCP 서버 상태 열기",
s11_app_c3: "변경사항 리뷰 모드 시작",
s11_app_c4: "스레드 ID/컨텍스트/레이트리밋 확인",
s11_app_c5: "피드백 다이얼로그 열기",
s11_app_note: "App 문서 기준으로 명령은 환경/권한에 따라 달라질 수 있으며, 활성화된 skill은 슬래시 목록에 추가될 수 있습니다.",
s11_refs: '공식 문서: <a href="https://developers.openai.com/codex/cli/slash-commands/" target="_blank">CLI slash commands</a> · <a href="https://developers.openai.com/codex/ide/slash-commands" target="_blank">IDE slash commands</a> · <a href="https://developers.openai.com/codex/app/commands" target="_blank">App commands</a>',
s11_c1: "모델 및 추론 수준 변경", s11_c2: "권한/승인 정책 변경", s11_c3: "채팅/에이전트 모드 전환",
s11_c10: "MCP 서버 추가/관리", s11_c4: "컨텍스트 압축", s11_c5: "git diff 보기", s11_c6: "스레드 ID/컨텍스트/사용량 확인",
s11_c7: "작업 리뷰 실행", s11_c8: "앱 사용 중 문제/버그 리포트", s11_c9: "세션 종료",
s11_k1: "입력창이 비었을 때 이전 사용자 메시지 편집", s11_k2: "세션 종료 (/exit)", s11_k3: "프롬프트 에디터 열기 (VISUAL/EDITOR)",
// Section 11
s12_title: "AGENTS.md 설정 가이드",
s12_intro: "Codex에게 프로젝트 맥락과 규칙을 알려주는 핵심 파일. CLI, App, IDE, Web 모두에서 참조합니다.",
s12_principles: "작성 5원칙",
s12_p1: "전문가 역할 지정", s12_p2: "코드로 보여주기", s12_p3: "구체적 명령",
s12_p4: "Always / Ask / Never", s12_p5: "리빙 다큐먼트 (지속 업데이트)",
s12_c1_t: "AGENTS.md에 꼭 넣을 것",
s12_c1_d: "repo layout, 실행 방법, build/test/lint 명령, 엔지니어링 규칙, 금지사항, 완료 기준과 검증 방법을 적어두면 반복 설명이 크게 줄어듭니다.",
s12_c2_t: "가까운 규칙이 우선합니다",
s12_c2_d: "개인 기본값은 전역 파일에, 팀 표준은 저장소 루트에, 하위 디렉토리의 로컬 규칙은 더 가까운 AGENTS.md에 두는 구조가 가장 관리하기 쉽습니다.",
s12_c3_t: "짧고 실용적으로 유지",
s12_c3_d: "모호한 규칙을 길게 적기보다 실제로 반복되는 실수만 추가하세요. 문서가 커지면 계획, 리뷰, 아키텍처 규칙은 별도 markdown으로 분리하는 편이 낫습니다.",
s12_callout_t: "같은 실수는 문서로 흡수하세요",
s12_callout_d: "Codex가 같은 실수를 두 번 반복하면 retrospective를 시키고 AGENTS.md를 업데이트하세요. 일회성 프롬프트보다 지속적인 품질 개선 효과가 큽니다.",
s12_refs: '출처: <a href="https://developers.openai.com/codex/learn/best-practices/" target="_blank">Codex Best practices</a> · <a href="https://developers.openai.com/codex/guides/agents-md" target="_blank">AGENTS.md guide</a>',
// Section 12
s13_title: "config.toml 설정",
s13_callout_t: "권장 설정 계층",
s13_callout_d: '개인 기본값은 <code>~/.codex/config.toml</code>, 저장소별 동작은 <code>.codex/config.toml</code>, 일회성 실험은 CLI override로 분리하는 패턴이 가장 안정적입니다. 프로젝트 설정은 저장소를 신뢰할 때만 로드되며, provider/auth/profile/telemetry 라우팅처럼 기기·계정에 묶인 키는 user-level config에 둬야 합니다.',
s13_beginner_t: "처음 사용자라면 이렇게 이해하면 됩니다",
s13_beginner_d: '<code>model</code>은 기본 두뇌, <code>review_model</code>은 리뷰 전용 두뇌, <code>sandbox_mode</code>는 작업 범위, <code>approval_policy</code>는 언제 멈춰서 묻는지, <code>web_search</code>는 웹 최신성 모드, <code>personality</code>는 기본 말투, <code>service_tier</code>는 속도 우선 경로, <code>default_permissions</code>는 기본 권한 묶음, <code>model_instructions_file</code>은 항상 따라야 할 개인 기본 지침입니다.',
s13_c1_t: "무엇을 고정할까",
s13_c1_d: "모델, model_provider, review_model, reasoning effort, sandbox mode, approval policy, web_search, profile, MCP 서버처럼 세션마다 흔들리면 품질 차이가 큰 항목부터 기본값으로 고정하세요. Amazon Bedrock 경로를 쓰는 팀은 AWS credential, region, model provider 설정도 운영 표준에 포함해야 합니다. Windows 네이티브 운영에서는 `windows.sandbox`, `windows.sandbox_private_desktop`을 함께 검토하고, remote connection은 App의 Settings > Connections와 workspace Remote Control 권한에서 관리하세요.",
s13_c2_t: "권한은 좁게 시작",
s13_c2_d: "처음에는 기본 권한과 보수적인 sandbox/approval 조합으로 시작하고, 신뢰할 수 있는 저장소나 반복된 워크플로에만 점진적으로 권한을 넓히는 편이 안전합니다. 세밀한 제어가 필요하면 현재 config reference에 설명된 `approval_policy = { granular = { ... } }` 형태를 검토하세요.",
s13_c3_t: "설정 문제를 먼저 의심",
s13_c3_d: "품질 문제가 보이면 프롬프트보다 먼저 작업 디렉토리, 쓰기 권한, 기본 모델, 필요한 도구 연결이 맞는지 확인하세요. 많은 실패는 setup mismatch에서 나옵니다.",
s13_c4_t: "팀 관리자는 requirements.toml도 확인",
s13_c4_d: "Business/Enterprise 운영자는 admin-enforced requirements로 approval, sandbox, web_search, MCP allowlist, Browser Use, Computer Use, in-app browser, sandboxed network policy 같은 기능을 사용자 설정보다 우선 적용할 수 있습니다.",
s13_refs: '출처: <a href="https://developers.openai.com/codex/learn/best-practices/" target="_blank">Codex Best practices</a> · <a href="https://developers.openai.com/codex/config-reference/" target="_blank">config reference</a> · <a href="https://developers.openai.com/codex/concepts/sandboxing/" target="_blank">sandboxing</a> · <a href="https://developers.openai.com/codex/amazon-bedrock" target="_blank">Codex on Amazon Bedrock</a> · <a href="https://developers.openai.com/api/docs/guides/amazon-bedrock" target="_blank">OpenAI models in Amazon Bedrock</a> · <a href="https://openai.com/index/openai-frontier-models-and-codex-are-now-available-on-aws/" target="_blank">OpenAI and Codex on AWS</a>',
// Section 13
s14_title: "MCP (Model Context Protocol)",
s14_intro: "MCP는 Codex가 외부 도구와 문서를 읽고, 브라우저를 제어하고, 원격 시스템과 연결할 수 있게 해주는 표준 프로토콜입니다.",
s14_beginner_title: "처음이라면 Docs MCP부터 연결하세요",
s14_beginner_desc: "Codex는 MCP 없이도 사용할 수 있습니다. 하지만 처음 시작하는 사람이라면 OpenAI Docs MCP 하나만 연결해도 설치, 모델, 명령어, API 문서를 훨씬 정확하게 찾게 됩니다.",
s14_step1: "OpenAI Docs MCP를 추가합니다.",
s14_step2: "codex mcp list / get으로 연결 상태를 확인합니다.",
s14_step3: "AGENTS.md에 문서 MCP를 우선 사용하라는 문장을 넣습니다.",
s14_step4: "이후 OpenAI 관련 질문을 문서 기반으로 물어봅니다.",
s14_agents_title: "AGENTS.md에 넣으면 좋은 문장",
s14_agents_desc: "이 문장이 없으면 Codex에게 매번 문서 MCP를 명시적으로 사용하라고 요청해야 할 수 있습니다.",
s14_prompt_title: "연결 후 바로 써볼 질문",
s14_prompt_desc: "MCP가 실제로 동작하는지 확인하려면, 최신 OpenAI 문서에서 특정 항목을 찾아 요약하게 해보는 것이 가장 빠릅니다.",
s14_scope_title: "설정 위치와 범위",
s14_scope_desc: "CLI와 IDE Extension은 MCP 설정을 공유합니다. 전역은 ~/.codex/config.toml, 프로젝트별로만 쓰고 싶다면 ./.codex/config.toml을 사용할 수 있습니다. 더 엄격하게 관리하려면 enabled, required, enabled_tools, disabled_tools 같은 옵션도 설정할 수 있습니다.",
s14_c1_t: "언제 MCP를 써야 할까",
s14_c1_d: "필요한 맥락이 저장소 밖에 있거나, 데이터가 자주 바뀌거나, 붙여넣기 대신 도구 호출로 재현 가능한 흐름을 만들고 싶을 때 MCP가 가장 효과적입니다.",
s14_c2_t: "지원 방식",
s14_c2_d: "Codex는 STDIO 서버와 OAuth를 포함한 Streamable HTTP 서버를 지원합니다. 먼저 설치가 쉬운 서버부터 연결해 반복 루프를 줄이세요.",
s14_c3_t: "도구는 적게 시작",
s14_c3_d: "처음부터 모든 툴을 연결하지 말고, 지금 수동으로 자주 반복하는 작업 하나나 둘만 줄여주는 도구부터 붙이세요. 신뢰가 생긴 뒤 확장하는 편이 낫습니다.",
th_server: "서버", th_purpose: "용도", th_package: "추천 순서",
mcp_docs_purpose: "OpenAI 공식 개발자 문서를 검색하고 읽기",
mcp_docs_priority: "가장 먼저",
mcp_playwright_purpose: "브라우저 열기, 클릭, UI 상태 확인",
mcp_playwright_priority: "UI 테스트가 필요할 때",
mcp_github_purpose: "원격 PR, 이슈, 저장소 메타데이터 다루기",
mcp_github_priority: "Git만으로 부족할 때",
mcp_figma_purpose: "디자인 파일과 컴포넌트 스펙 확인",
mcp_figma_priority: "디자인 협업이 많을 때",
s14_refs: '출처: <a href="https://developers.openai.com/learn/docs-mcp" target="_blank">Docs MCP</a> · <a href="https://developers.openai.com/codex/learn/best-practices/" target="_blank">Codex Best practices</a> · <a href="https://developers.openai.com/codex/mcp" target="_blank">Codex MCP</a>',
// Section 14
s15_title: "세션 관리",
s15_callout_t: "어려운 작업은 plan-first로 시작",
s15_callout_d: "복잡하거나 모호한 작업은 바로 코딩하지 말고 먼저 계획을 세우게 하세요. Plan mode, 인터뷰식 요구사항 정리, 실행 계획 템플릿은 긴 작업의 실패율을 줄여줍니다.",
s15_c1_t: "스레드는 작업 단위로",
s15_c1_d: "프로젝트 전체를 하나의 스레드로 몰지 말고, 하나의 문제를 푸는 동안만 같은 스레드를 유지하세요. 일이 실제로 갈라질 때만 fork하는 편이 문맥 품질이 좋습니다.",
s15_c2_t: "컨텍스트는 milestone 기준으로 정리",
s15_c2_d: "resume으로 이어가고, compact는 milestone 이후에만 쓰세요. 너무 이른 요약은 추론 흔적과 결정 근거를 지워 품질을 떨어뜨릴 수 있습니다.",
s15_c3_t: "병렬 agent는 범위가 잘린 일에만",
s15_c3_d: "메인 스레드는 요구사항과 최종 결정을 맡기고, 탐색·테스트·트리아지처럼 읽기 중심의 bounded work만 병렬 agent나 subagent로 분리하세요. 쓰기 충돌이 나는 병렬 편집은 더 조심해야 합니다.",
s15_refs: '출처: <a href="https://developers.openai.com/codex/learn/best-practices/" target="_blank">Codex Best practices</a> · <a href="https://developers.openai.com/codex/cli/slash-commands/" target="_blank">CLI slash commands</a> · <a href="https://developers.openai.com/codex/concepts/multi-agents/" target="_blank">Multi-agents concepts</a> · <a href="https://developers.openai.com/cookbook/articles/codex_exec_plans" target="_blank">Execution plans guide</a>',
// Section 15
s16_title: "비대화형 자동화 & CI/CD",
s16_cloud_title: "Codex Web 자동 리뷰",
s16_cloud_desc: 'GitHub 연결 + 리뷰 대상 저장소 선택 후 "Automatic reviews" 활성화 → 새 PR마다 자동 리뷰, AGENTS.md 규칙 기반 커스터마이징.',
s16_c1_t: "신뢰할 수 있는 변경 루프",
s16_c1_d: "변경만 시키지 말고 테스트 작성 또는 수정, 적절한 체크 실행, 동작 확인, diff 리뷰까지 같이 요구하세요. Codex는 코드 생성보다 검증 루프까지 맡길 때 더 유용합니다.",
s16_c2_t: "리뷰도 규칙화할 수 있습니다",
s16_c2_d: "base branch 기준 PR 스타일 리뷰, 미커밋 변경 리뷰, 특정 커밋 리뷰를 `/review`로 돌릴 수 있고, `code_review.md`를 AGENTS.md에서 참조하면 팀 기준을 재사용할 수 있습니다.",
s16_c3_t: "자동화 후보를 고르는 법",
s16_c3_d: "최근 커밋 요약, 버그 스캔, 릴리스 노트 초안, CI 실패 점검, 스탠드업 요약처럼 이미 수동으로 안정화된 반복 작업부터 자동화하세요. 같은 스레드를 이어야 하면 thread automation, 매번 새로 시작해야 하면 standalone automation이 맞고, 프로젝트 스코프 자동화는 앱이 실행 중이며 대상 프로젝트가 디스크에 있어야 합니다. unattended 자동화는 기본 sandbox를 그대로 쓰고, 조직 정책이 허용하면 `approval_policy = \"never\"`를 우선 사용합니다.",
s16_callout_t: "skill이 방법을, automation이 일정을 정의합니다",
s16_callout_d: "아직 사람의 조향이 많이 필요한 흐름이면 먼저 skill로 묶고, 예측 가능해진 뒤 자동화로 옮기세요. 너무 이른 자동화는 품질 편차만 키웁니다.",
s16_refs: '출처: <a href="https://developers.openai.com/codex/learn/best-practices/" target="_blank">Codex Best practices</a> · <a href="https://developers.openai.com/codex/app/review" target="_blank">Codex app review</a> · <a href="https://developers.openai.com/codex/integrations/github" target="_blank">GitHub integration</a> · <a href="https://developers.openai.com/codex/app/automations" target="_blank">Automations</a> · <a href="https://developers.openai.com/codex/app/worktrees" target="_blank">Worktrees</a> · <a href="https://openai.com/index/dell-codex-enterprise-partnership/" target="_blank">Dell enterprise partnership</a> · <a href="https://openai.com/index/gartner-2026-agentic-coding-leader/" target="_blank">Gartner enterprise coding agents</a>',
// Section 16
s17_title: "Prompting Codex Agents",
s17_credit: "이제 프롬프트는 문장 요령보다 작업 규칙을 분명히 정하는 일이 더 중요합니다.",
s17_intro_title: "팁 모음에서 작업 규칙으로",
s17_intro_desc: "GPT-5.5 기준의 핵심은 세세한 절차를 모두 적는 것보다 목표 결과, 성공 기준, 제약, 사용할 수 있는 근거를 분명히 주고 모델이 적절한 경로를 고르게 하는 데 있습니다.",
s17_intro_ref: '레퍼런스: <a href="https://developers.openai.com/codex/learn/best-practices/" target="_blank">Codex Best practices</a> · <a href="https://developers.openai.com/codex/use-cases/follow-goals" target="_blank">Follow goals</a> · <a href="https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes" target="_blank">Enterprise/Edu release notes</a> · <a href="https://developers.openai.com/api/docs/guides/prompt-guidance?model=gpt-5.5" target="_blank">GPT-5.5 Prompt guidance</a>',
s17_first_use_t: "Strong first use: 결과와 멈춤 기준을 먼저 주세요",
s17_first_use_d: "큰 저장소나 중요한 작업일수록 Goal, Context, Constraints, Success criteria, Stop rules를 짧게 주면 Codex가 불필요한 루프 없이 더 검증 가능한 결과를 냅니다.",
s17_reasoning_t: "reasoning은 난이도 기준으로 고르세요",
s17_reasoning_d: "작고 명확한 작업은 낮은 설정에서 시작하고, 복잡한 변경과 디버깅은 medium/high로 올립니다. 품질이 부족할 때는 reasoning을 올리기 전에 완료 기준, 검증, 근거 규칙을 먼저 점검하세요.",
s17_g1: "Goal: 무엇을 바꾸거나 만들고 싶은지 한 문장으로 못 박습니다.",
s17_g2: "Context: 관련 파일, 폴더, 문서, 에러 로그, 예시를 함께 지정합니다.",
s17_g3: "Constraints: 아키텍처, 안전 기준, 코딩 컨벤션, 금지 규칙을 분명히 합니다.",
s17_g4: "Success + Stop rules: 테스트, 기대 동작, 필요한 근거, 멈출 조건을 함께 명시합니다.",
s17_c1_t: "1. Outcome-first contract",
s17_c1_d: "절차를 길게 잠그기보다 목표 결과, 필수 출력 필드, 성공 기준, 부족한 근거가 있을 때의 행동을 먼저 고정하세요.",
s17_c2_t: "2. Follow-through",
s17_c2_d: "되돌릴 수 있고 저위험인 다음 단계는 자동으로 진행하고, 외부 부작용이나 파괴적 변경만 별도 승인을 받게 해야 합니다.",
s17_c3_t: "3. Tool Loops",
s17_c3_d: "정확성이 중요하면 도구를 한 번 쓰고 멈추지 않게 하고, 선행 조회, 빈 결과 복구, 검색 예산, 완료 판정까지 묶어줘야 합니다.",
s17_c4_t: "4. Reasoning Last",
s17_c4_d: "품질이 부족할 때 무조건 reasoning effort부터 올리기보다, 먼저 요청 방식과 검증 규칙을 보강하는 것이 더 효율적입니다.",
s17_link1_kicker: "Shortcuts",
s17_link1_title: "챕터 바로가기",
s17_link1_desc: "상단 점프 바에서 설치, 실행 규칙, 비교, FAQ 같은 다른 챕터로 바로 이동할 수 있습니다.",
s17_link1_cta: "상단 점프 바로",
s17_link2_kicker: "Official",
s17_link2_title: "OpenAI Prompt Guidance",
s17_link2_desc: "GPT-5.5의 outcome-first prompting, reasoning effort, phase, compaction, validation, research/citation 규칙까지 포함한 공식 기준 문서를 같이 보세요.",
s17_link2_cta: "공식 문서 보기",
pg_hero_badge: "Deep guide",
pg_hero_title: "Prompting modern Codex agents",
pg_hero_sub: "긴 작업, 도구 사용, 리서치, 코딩 에이전트에 맞게 출력 형식과 검증 방식을 설계하는 실전 가이드입니다.",
pg_hero_cta: "가이드 펼쳐보기",
pg_hero_back: "Codex 101으로 돌아가기",
pg_hero_meta_sections: "6개 섹션",
pg_hero_meta_note: "좌측 탭이 스크롤 중에도 계속 보여서 길을 잃지 않게 합니다.",
pg_nav_kicker: "이 페이지",
pg_nav_back: "Codex 101 섹션 17",
pg_nav_official: "공식 Prompt Guidance",
pg_toc_title: "프롬프팅 맵",
pg_toc_lead: "이 페이지는 GPT-5.5 Prompt Guidance와 최신 Codex best practices를 바탕으로 실전용 프롬프팅 원칙을 재구성한 deep guide입니다.",
toc_prompt_1_t: "Behavior",
toc_prompt_1_d: "최근 Codex 모델이 어디에서 강하고, 어디에서 여전히 명시적 prompting이 필요한지 봅니다.",
toc_prompt_2_t: "핵심 요청 규칙",
toc_prompt_2_d: "출력 형식, 끝까지 진행하는 기준, 지시 우선순위를 짧은 블록으로 정리합니다.",
toc_prompt_3_t: "Agent Loops",
toc_prompt_3_d: "도구를 끝까지 쓰는 기준, 의존성 확인, 완료 기준, 검증 방식을 설계합니다.",
toc_prompt_4_t: "Specialized Workflows",
toc_prompt_4_d: "research, citations, coding agents, strict formats, customer writing에 맞는 패턴입니다.",
toc_prompt_5_t: "Runtime and Tuning",
toc_prompt_5_d: "reasoning_effort, phase, compaction, migration 순서를 다룹니다.",
toc_prompt_6_t: "Starter Playbooks",
toc_prompt_6_d: "바로 가져다 쓸 수 있는 coding/research prompt skeleton입니다.",
pg_s1_title: "무엇이 달라졌나",
pg_s1_credit: "최신 Codex best practices를 Codex 워크플로 기준으로 재정리했습니다.",
pg_s1_intro: "GPT-5.5 프롬프팅은 모든 단계를 과하게 지정하는 방식보다, 원하는 결과와 성공 기준을 선명하게 두고 필요한 도구·추론·검증 경로를 모델이 고르게 하는 쪽에 가깝습니다.",
pg_s1_f1: "긴 작업에서는 짧은 preamble으로 첫 단계와 진행 방향을 보여주면 체감 응답성이 좋아집니다.",
pg_s1_f2: "출력 형식은 잘 따르지만, 무거운 구조는 필요한 결과물이나 UI 안정성이 있을 때만 쓰는 편이 좋습니다.",
pg_s1_f3: "검색과 인용이 필요한 답변은 retrieval budget, 충분한 근거 기준, 근거가 없을 때의 행동을 같이 정해야 합니다.",
pg_s1_f4: "검증 가능한 작업은 테스트, lint, build, 렌더 확인처럼 실제 확인 도구를 프롬프트에 포함할수록 안정적입니다.",
pg_s1_callout_t: "목표를 먼저, 절차는 필요한 만큼만",
pg_s1_callout_d: "Goal, success criteria, constraints, output, stop rules를 짧게 고정하고, incomplete execution, format drift, 약한 citation처럼 관측한 실패를 고칠 때만 추가 블록을 붙이세요.",
pg_s2_title: "핵심 요청 규칙",
pg_s2_intro: "가장 큰 개선 효과는 목표 결과, 성공 기준, 자율 진행 기준, 충돌하는 지시의 우선순위를 짧고 명확한 규칙으로 적는 데서 나옵니다.",
pg_s2_c1_t: "Outcome contract",
pg_s2_c1_d: "무엇을 달성해야 하는지, 성공은 무엇으로 판단하는지, 필수 출력 필드는 무엇인지 먼저 고정하세요.",
pg_s2_c2_t: "Default follow-through",
pg_s2_c2_d: "저위험이고 되돌릴 수 있는 다음 단계는 진행하고, 파괴적 변경이나 외부 부작용만 승인을 받게 하세요.",
pg_s2_c3_t: "Instruction priority",
pg_s2_c3_d: "새 사용자 지시는 예전 스타일보다 우선하되, 안전·정직·권한 규칙은 예외로 두는 식으로 우선순위를 분명히 합니다.",
pg_s2_c4_t: "Scoped updates",
pg_s2_c4_d: "중간에 지시가 바뀌면 이번 응답에만 적용되는지, 남은 세션 전체에 적용되는지 범위를 적어 주세요.",
pg_s2_note: "모든 블록을 항상 넣지 말고, 실제로 실패를 줄여주는 조합만 남기는 편이 좋습니다.",
pg_s3_title: "Agent Loops",
pg_s3_intro: "긴 작업에서 흔한 실패는 “대충 끝난 것처럼 보이지만 실제로는 덜 끝난 상태”입니다. 도구를 어디까지 계속 쓸지, 빈 결과를 어떻게 복구할지, 결과를 어떻게 검증할지를 같이 정해야 합니다.",
pg_s3_c1_t: "Tool persistence",
pg_s3_c1_d: "정확도와 완결성에 도움이 되면 도구를 계속 쓰고, 한 번의 조회 결과만으로 너무 빨리 멈추지 않게 합니다.",
pg_s3_c2_t: "Dependency checks",
pg_s3_c2_d: "최종 행동이 명확해 보여도 선행 조회나 prerequisite step이 필요한지 먼저 확인해야 합니다.",
pg_s3_c3_t: "Selective parallelism",
pg_s3_c3_d: "독립적인 조회만 병렬화하고, 앞 단계 결과가 다음 행동을 바꾸는 경우는 순차 실행으로 남겨야 합니다.",
pg_s3_c4_t: "완료 기준",
pg_s3_c4_d: "배치, 목록, 페이지네이션 작업이라면 예상 범위와 처리 상태를 내부적으로 추적하도록 요구하세요.",
pg_s3_c5_t: "Verification loop",
pg_s3_c5_d: "끝나기 전에 요구사항 커버리지, grounding, 포맷, 안전성, 되돌리기 어려운 부작용 여부를 한 번 더 확인하게 하세요.",
pg_goals_t: "Goal mode는 장기 작업의 체크포인트 계약입니다",
pg_goals_d: "공식 Follow goals 문서와 2026년 5월 21일 Enterprise/Edu 릴리스 노트 기준으로 Goal mode는 Codex app, IDE extension, CLI에서 일반 제공됩니다. 한 번의 프롬프트보다 길지만 무한 backlog는 아닌 작업에 쓰고, 시작할 때 목표·완료 조건·검증 명령·중간 체크포인트·중단/승인 조건을 함께 적으세요.",
pg_goals_f1: "`/goal` 또는 Goal mode UI가 보이지 않으면 Codex app/IDE/CLI를 업데이트하고, Enterprise/Edu workspace에서는 관리자가 기능을 제한했는지 확인하세요.",
pg_goals_f2: "상태 업데이트는 “현재 checkpoint, 검증한 것, 남은 것, 막힌 것” 네 항목으로 짧게 요구하세요. 목표가 바뀌면 새 목표로 다시 계약을 잡는 편이 안전합니다.",
pg_goals_f3: "마이그레이션, 여러 파일 리팩터링, prototype polish, eval을 돌리며 prompt를 다듬는 작업처럼 완료 조건이 관찰 가능한 일에 특히 잘 맞습니다.",
pg_s4_title: "Specialized Workflows",
pg_s4_intro: "공식 가이드의 핵심은 모든 작업에 같은 프롬프트를 쓰지 말고, 작업 유형별로 요청 규칙을 다르게 잡으라는 점입니다.",
pg_s4_c1_t: "Research and citations",
pg_s4_c1_d: "검색과 합성이 필요한 작업은 sub-question 분해, retrieval budget, 재검색 조건, citation gating을 따로 줘야 합니다.",
pg_s4_c2_t: "Strict structured outputs",
pg_s4_c2_d: "JSON, SQL, XML처럼 파싱 민감한 출력은 “그 형식만 반환”과 마감 전 self-check를 함께 요구하세요.",
pg_s4_c3_t: "Coding agents",
pg_s4_c3_d: "셸 도구, 편집 도구, 사용자 업데이트 빈도, 검증 강도를 명확히 두면 코딩 에이전트의 일관성이 올라갑니다.",
pg_s4_c4_t: "Customer-facing writing",
pg_s4_c4_d: "성격(personality)은 세션 기본값으로, 채널/톤/길이 제한은 응답 단위 제어로 분리하는 편이 좋습니다.",
pg_s4_block1_t: "Research mode starter",
pg_s4_block2_t: "Coding agent starter",
pg_s5_title: "Runtime and Tuning",
pg_s5_intro: "reasoning effort는 첫 번째 해결책이 아니라 마지막 미세 조정 노브에 가깝습니다. 요청 방식과 검증 흐름을 먼저 손보는 편이 더 값집니다.",
pg_s5_th: "언제 여기서 시작할까",
pg_s5_r1: "짧은 변환, 실행 중심 작업, 빠른 tool routing, latency가 중요한 플로우",
pg_s5_r2: "약간의 해석은 필요하지만 속도도 중요한 작업",
pg_s5_r3: "다문서 합성, 충돌 해소, 정책/전략성 글쓰기, 복잡한 코딩 계획",
pg_s5_r4: "장기 에이전트 작업에서 eval로 분명한 이득이 있을 때만 선택",
pg_s5_f1: "Responses API의 <code>phase</code>는 preamble 같은 중간 업데이트와 final answer를 구분합니다. assistant item을 수동 재전송한다면 기존 <code>phase</code> 값을 그대로 보존해야 합니다.",
pg_s5_f2: "Compaction을 쓸 때는 milestone 이후에만 압축하고, 압축 전후 프롬프트의 기능적 의미를 바꾸지 않는 편이 안전합니다.",
pg_s5_f3: "GPT-5.5로 옮길 때는 모델만 먼저 바꾸고 reasoning을 고정한 뒤 eval로 차이를 보는 one-change-at-a-time 방식이 낫습니다.",
pg_s5_callout_t: "먼저 요청 방식, 그 다음 reasoning",
pg_s5_callout_d: "품질이 너무 문자 그대로이거나 중간에 멈춘다면 reasoning을 올리기 전에 완료 기준, 도구 사용 기준, 검증, 인용 규칙을 먼저 붙여 보세요.",
pg_s6_title: "Starter Playbooks",
pg_s6_intro: "아래 템플릿은 Codex 101에서 바로 가져다 쓸 수 있는 최소 골격입니다. outcome, success criteria, constraints, output, stop rules를 먼저 두고, 필요 없는 블록은 빼세요.",
pg_s6_block1_t: "Coding task prompt",
pg_s6_block2_t: "Research task prompt",
// Section 17
s18_title: "고급 활용 기법",
s18_intro: "최신 best practices와 customization 문서를 같이 보면 장기적으로 잘 쓰는 사용자는 반복 작업을 skill로 만들고, 팀에 공유할 때는 plugin으로 묶고, 외부 시스템 연결은 MCP로 붙이고, 장기 작업은 bounded delegation으로 나눕니다.",
s18_field_title: "커뮤니티 실전 팁: Gabriel Chua, VB, Tibo에게서 가져온 운영 습관",
s18_field_credit: "이번 실행에서는 Chrome에서 X/Threads/LinkedIn 본문을 읽을 수 있었고, X/LinkedIn에서 공식 발표와 일치하는 role-specific plugins, Sites, annotations 반복을 확인했습니다. Threads에서는 공식 문서로 반영할 새 claim을 찾지 못했으며, 검증되지 않은 소셜 claim은 추가하지 않았습니다.",
s18_field_c1_t: "Model + Harness + Surfaces로 사고하기",
s18_field_c1_d: "Gabriel Chua의 설명처럼 Codex를 모델 하나가 아니라 모델, 하네스, 표면의 결합으로 보면 업데이트를 더 정확히 읽을 수 있습니다. 모델이 바뀐 것인지, 도구/실행 루프가 바뀐 것인지, App/CLI/IDE 표면이 바뀐 것인지 나눠서 확인하세요.",
s18_field_c2_t: "다른 에이전트 옆에 Codex 리뷰어를 붙이기",
s18_field_c2_d: "VB가 공유한 Codex Plugin for Claude Code 흐름은 좋은 패턴입니다. 기본 변경에는 읽기 전용 review, 큰 설계 변경에는 adversarial review, 막힌 작업에는 rescue를 붙여 한 에이전트의 blind spot을 다른 에이전트가 보게 만드세요.",
s18_field_c3_t: "Subagent는 병렬 읽기 작업부터",
s18_field_c3_d: "Tibo가 소개한 subagents 흐름은 main thread를 깨끗하게 유지하는 데 특히 유용합니다. 보안, 테스트 공백, 유지보수성처럼 독립적으로 볼 수 있는 항목에 agent를 하나씩 붙이고, 최종 요약만 메인 스레드로 가져오세요.",
s18_field_c4_t: "Automation을 코딩 밖 업무까지 확장하기",
s18_field_c4_d: "VB의 9AM 자동화 사례처럼 Slack, Gmail, Calendar를 교차 확인해 회의 pre-brief를 만드는 흐름은 Codex를 단순 코딩 도구가 아니라 작업 허브로 쓰는 예입니다. 반복되는 준비 업무는 먼저 수동 skill로 안정화한 뒤 automation으로 올리세요.",
s18_field_c5_t: "권한을 무작정 열지 말고 위험별로 나누기",
s18_field_c5_d: "Tibo와 VB가 언급한 Guardian Approvals/approval 흐름의 핵심은 YOLO 모드가 아니라 위험한 tool call만 올리는 것입니다. auth, data loss, infra, rollback, race condition은 adversarial review와 명시적 승인 대상으로 분리하세요.",
s18_field_c6_t: "Windows에서도 App-first를 기본 경로로 보기",
s18_field_c6_d: "Tibo의 Windows 언급은 현재 공식 Windows App 문서 흐름과도 맞습니다. 기업/학교 환경에서는 허용해야 할 앱을 줄이는 것이 중요하므로, 네이티브 App + sandbox + 필요 시 WSL2 전환을 기본 운영안으로 잡는 편이 깔끔합니다.",
s18_field_refs: '출처: <a href="https://www.linkedin.com/pulse/how-i-think-codex-gabriel-chua-ukhic" target="_blank">Gabriel Chua, How I Think About Codex</a> · <a href="https://openai.com/index/unrolling-the-codex-agent-loop/" target="_blank">OpenAI, Unrolling the Codex agent loop</a> · <a href="https://github.com/openai/codex-plugin-cc" target="_blank">openai/codex-plugin-cc</a> · <a href="https://developers.openai.com/codex/subagents" target="_blank">Codex Subagents</a> · <a href="https://sub.thursdai.news/p/thursdai-ai-engineer-europe-mythos" target="_blank">ThursdAI notes with VB</a> · <a href="https://twstalker.com/thsottiaux" target="_blank">Tibo X mirror</a>',
s18_c1_t: "Skills 먼저, Plugins는 공유 단계",
s18_c1_d: "같은 프롬프트나 같은 수정 지시를 반복한다면 먼저 skill로 정리하세요. 그 workflow를 팀이나 여러 프로젝트에 배포하고 싶어지면 plugin으로 묶어 skill, 선택적 app 연동, MCP 설정을 한 번에 설치 가능하게 만드는 방식이 현재 공식 구조와 맞습니다.",
s18_c2_t: "병렬 agent는 읽기 중심 작업부터",
s18_c2_d: "메인 에이전트는 핵심 의사결정에 남기고, 탐색, 테스트, 로그 트리아지, 릴리스 초안 같은 보조 작업만 병렬화하면 결과를 통합하기 쉽습니다.",
s18_c3_t: "Plugin으로 묶기 좋은 반복 작업",
s18_c3_d: "로그 트리아지, PR 체크리스트 리뷰, 마이그레이션 계획, incident summary, 표준 디버깅 플로우처럼 skill 하나로 끝나지 않고 app 연동이나 MCP가 함께 필요한 workflow는 plugin으로 묶기 좋은 후보입니다.",
s18_callout_t: "로컬 skill로 검증하고 plugin으로 공유하세요",
s18_callout_d: "반복 작업이 아직 많이 흔들리면 automation보다 skill이 먼저입니다. 수동으로 안정화된 뒤 plugin으로 패키징하고, 그 다음 자동화를 얹는 순서가 가장 안전합니다.",
s18_m1: "지속 규칙을 프롬프트에만 길게 넣고 AGENTS.md나 skill로 옮기지 않는 실수",
s18_m2: "build/test 명령을 알려주지 않아 Codex가 자기 결과를 검증하지 못하게 하는 실수",
s18_m3: "복잡한 작업에서 planning을 건너뛰고 바로 구현부터 시작하는 실수",
s18_m4: "워크트리 없이 같은 파일에 여러 live thread를 동시에 붙이는 실수",
s18_m5: "프로젝트 하나에 스레드 하나만 유지해 문맥을 비대하게 만드는 실수",
s18_refs: '출처: <a href="https://developers.openai.com/codex/learn/best-practices/" target="_blank">Codex Best practices</a> · <a href="https://developers.openai.com/codex/concepts/customization/" target="_blank">Customization</a> · <a href="https://developers.openai.com/codex/skills" target="_blank">Skills</a> · <a href="https://developers.openai.com/codex/plugins/" target="_blank">Plugins</a> · <a href="https://developers.openai.com/codex/concepts/multi-agents/" target="_blank">Multi-agents</a> · <a href="https://openai.com/index/building-self-improving-tax-agents-with-codex/" target="_blank">Self-improving tax agents with Codex</a>',
// Section 19
s20_title: "자주 묻는 질문 (FAQ)",
faq1_q: "Codex CLI / App / IDE / Web의 차이는?",
faq1_a: "모두 같은 Codex App Server 프로토콜을 사용하지만 인터페이스가 다릅니다. CLI는 터미널, App은 macOS/Windows 데스크톱 멀티에이전트, IDE는 VS Code/JetBrains 통합, Web은 브라우저에서 GitHub 연동.",
faq2_q: "무료로 사용할 수 있나요?",
faq2_a: "네. 현재 공식 문서는 모든 ChatGPT 플랜에 Codex가 포함된다고 설명합니다. 다만 정확한 사용 한도와 rate limit은 모델, 작업 크기, 로컬/클라우드 실행 여부에 따라 달라지므로 실제 사용 전에는 공식 가격표와 usage dashboard를 함께 확인하는 편이 좋습니다.",
faq3_q: "Full Auto 모드가 안전한가요?",
faq3_a: "OS 수준 샌드박싱 + 네트워크 차단이 적용됩니다. Git으로 관리하면 언제든 되돌릴 수 있습니다.",
faq4_q: "GPT-5.3-Codex-Spark는 뭔가요?",
faq4_a: "2026년 2월 12일 공개된 텍스트 중심 초고속 코딩 모델입니다. ChatGPT Pro용 research preview가 기본 경로이고, Pricing 문서도 API launch 시점에는 제공되지 않는다고 설명합니다. 발표문은 소수 디자인 파트너 대상 제한적 API 테스트와 별도 사용 한도를 함께 언급합니다.",
faq5_q: "오프라인에서 사용 가능한가요?",
faq5_a: "Codex CLI는 오픈소스(Rust)이며, 타사 모델 프로바이더(Ollama 등)를 연결할 수 있지만, 품질은 OpenAI 모델보다 떨어집니다.",
// Section 20
s21_title: "참고 자료",
s21_prompt_guide: "Prompting Codex Agents 섹션",
field_tips_nav: "OpenAI 담당자 팁",
field_tips_title: "OpenAI 담당자들이 전하는 팁",
field_tips_intro: "OpenAI Developer Experience와 Codex 담당자들의 공개 글, X 원글/미러, 인터뷰, 공식 저장소를 대조해 실무자가 바로 적용할 수 있는 운영 습관만 추렸습니다.",
field_tip_1_t: "Codex를 Model + Harness + Surfaces로 나눠 읽기",
field_tip_1_d: "새 모델 발표만 보지 말고, 모델 자체가 바뀐 것인지, 하네스의 도구 실행/검증 루프가 바뀐 것인지, App·CLI·IDE·Web 표면이 바뀐 것인지 분리해서 보세요. 업데이트를 훨씬 덜 헷갈리게 읽을 수 있습니다.",
field_tip_2_t: "다른 에이전트 옆에 Codex 리뷰어 붙이기",
field_tip_2_d: "Codex Plugin for Claude Code 흐름처럼, 한 에이전트가 만든 변경을 Codex가 읽기 전용 리뷰·adversarial review·rescue 모드로 다시 보게 하세요. 큰 변경일수록 두 번째 관점이 버그와 설계 공백을 잘 잡습니다.",
field_tip_3_t: "Subagent는 읽기 중심 병렬 작업부터",
field_tip_3_d: "보안, 테스트 공백, 유지보수성처럼 독립적으로 검토할 수 있는 항목에 subagent를 붙이면 main thread가 깔끔하게 유지됩니다. 처음에는 수정 작업보다 조사·리뷰·요약 작업에 쓰는 편이 안정적입니다.",
field_tip_4_t: "Vibe coding보다 AI teammate + PR review",
field_tip_4_d: "인터뷰에서 반복되는 메시지는 단순 자동완성보다 협업 방식입니다. Codex를 PR을 설명하고, 테스트를 돌리고, 리뷰 가능한 단위로 변경을 나누는 팀원처럼 다루면 결과 품질이 안정됩니다.",
field_tip_5_t: "작업을 검증 가능하게 만들기",
field_tip_5_d: "잘 되는 에이전트 작업은 “해줘”보다 완료 조건이 분명합니다. 실행할 명령, 확인할 화면, 실패 시 멈출 기준, 출처를 남기는 방식을 프롬프트에 넣으면 Codex가 더 좋은 루프를 만듭니다.",
field_tip_6_t: "Codex를 하나의 command center로 보기",
field_tip_6_d: "Developer Experience 쪽 사례들은 Codex를 코드 생성기 하나로 보지 않습니다. Slack, Gmail, Calendar, 브라우저, 문서, 저장소를 연결해 반복되는 준비·조사·리뷰 업무까지 묶는 운영 허브로 보는 편이 맞습니다.",
pet_recording_kicker: "Codex Pet 실전 녹화본",
pet_recording_title: "왜 Codex Pet이 나왔는지 보기",
pet_recording_desc: "Codex Pet은 단순한 캐릭터 예제가 아니라, `hatch-pet` 같은 skill을 설치해 base art 생성, 8x9 spritesheet/atlas 조립, QA contact sheet, preview, `pet.json` 패키징까지 반복 가능한 절차로 실행하는 사례입니다. 이 영상은 설명을 덧붙인 튜토리얼이 아니라 실제 제작 과정을 풀로 기록한 녹화본입니다.",
pet_recording_cta: "YouTube에서 풀 녹화본 보기",
field_tips_sources_label: "원글 및 근거",
field_tips_refs: '원글 및 근거: <a href="https://x.com/gabrielchua" target="_blank">Gabriel Chua X</a> · <a href="https://www.linkedin.com/pulse/how-i-think-codex-gabriel-chua-ukhic" target="_blank">How I Think About Codex</a> · <a href="https://x.com/reach_vb" target="_blank">VB X</a> · <a href="https://x.com/reach_vb/status/2039251986357338257" target="_blank">VB Codex plugin post</a> · <a href="https://github.com/openai/codex-plugin-cc" target="_blank">openai/codex-plugin-cc</a> · <a href="https://developers.openai.com/codex/skills" target="_blank">Codex Skills</a> · <a href="https://x.com/thsottiaux" target="_blank">Tibo X</a> · <a href="https://developers.openai.com/codex/subagents" target="_blank">Codex Subagents</a> · <a href="https://x.com/kagigz" target="_blank">Katia X</a> · <a href="https://www.computingdeutschland.de/interview/2025/developer-experience-mit-openai" target="_blank">Katia DevEx interview</a> · <a href="https://x.com/dkundel/status/2018436269907603590" target="_blank">Dominik Kundel X post</a> · <a href="https://vivatech.com/speakers/e5bb6392-2f32-f011-8b3d-6045bd903b46/" target="_blank">Romain Huet profile</a> · <a href="https://openai.com/index/unrolling-the-codex-agent-loop/" target="_blank">Unrolling the Codex agent loop</a> · <a href="https://youtu.be/l-eXsCldCgQ" target="_blank">Codex Pet full recording</a>',
// Footer
footer_contrib: '이 가이드는 오픈소스입니다. 오류 수정, 번역 개선, 새 팁 추가 등 <a href="https://github.com/swhan0329/codex-101" target="_blank">PR을 환영합니다</a>.',
footer_date_prefix: "최종 업데이트",
// Code blocks (Korean)
code_s5_install: '<code><span class="c"># npm</span>\nnpm install -g @openai/codex\n\n<span class="c"># Homebrew (macOS)</span>\nbrew install codex\n\n<span class="c"># 업데이트</span>\nnpm update -g @openai/codex</code>',
code_s5_auth: '<code><span class="c"># ChatGPT 계정 로그인 (권장)</span>\ncodex login\n\n<span class="c"># 또는 API 키 사용</span>\nexport OPENAI_API_KEY="your-api-key-here"</code>',
code_s6_usage: '<code><span class="c"># 인터랙티브 모드</span>\ncodex\n\n<span class="c"># 프롬프트와 함께</span>\ncodex "explain this codebase"\n\n<span class="c"># 특정 디렉토리</span>\ncodex --cd /path/to/project\n\n<span class="c"># 파일 참조</span>\n@src/components/Button.tsx 리팩토링해줘\n\n<span class="c"># 쉘 명령어 실행</span>\n!npm test</code>',
code_s12_loc: '<code><span class="c"># AGENTS.md 위치 및 우선순위</span>\nAGENTS.override.md → 최우선\n./AGENTS.md → 가장 가까운 파일\n../AGENTS.md → 상위 디렉토리\n~/.codex/AGENTS.md → 전역 개인 설정</code>',
code_s12_init: '<code><span class="c"># 자동 생성</span>\n/init</code>',
code_s13_config: '<code><span class="c">#:schema https://developers.openai.com/codex/config-schema.json</span>\n<span class="c"># ~/.codex/config.toml</span>\nmodel = "gpt-5.5"\nreview_model = "gpt-5.4"\nmodel_provider = "openai"\napproval_policy = "on-request"\nsandbox_mode = "workspace-write"\nweb_search = "cached"\nmodel_reasoning_effort = "medium"\nplan_mode_reasoning_effort = "medium"\npersonality = "friendly"\nservice_tier = "fast"\ndefault_permissions = "project_safe"\nmodel_instructions_file = "~/.codex/instructions.md"\n\n[sandbox_workspace_write]\nnetwork_access = false\n\n[tools]\nweb_search = { context_size = "medium", allowed_domains = ["developers.openai.com", "openai.com"] }\nview_image = true\n\n[permissions.project_safe.filesystem]\n":project_roots" = "write"\n\n[permissions.project_safe.network]\nenabled = false\n\n[profiles.fast]\nmodel = "gpt-5.4-mini"\nmodel_reasoning_effort = "low"\n\n[profiles.deep]\nmodel = "gpt-5.5"\nmodel_reasoning_effort = "high"\n\n[windows]\nsandbox = "elevated"\nsandbox_private_desktop = true\n\n<span class="c"># 필요한 경우 세밀 승인 정책</span>\n# approval_policy = { granular = {\n# sandbox_approval = true,\n# rules = true,\n# mcp_elicitations = true,\n# request_permissions = false,\n# skill_approval = false\n# } }\n\n<span class="c"># OpenAI Docs MCP</span>\n[mcp_servers.openaiDeveloperDocs]\nurl = "https://developers.openai.com/mcp"\nrequired = true</code>',
code_s14_mcp: '<code><span class="c"># 1) 초보자 추천: OpenAI Docs MCP</span>\ncodex mcp add openaiDeveloperDocs --url https://developers.openai.com/mcp\n\n<span class="c"># 2) 연결 확인</span>\ncodex mcp list\ncodex mcp get openaiDeveloperDocs\n\n<span class="c"># 3) UI 작업이 많다면 Playwright도 추가</span>\ncodex mcp add playwright -- npx --yes @playwright/mcp@latest</code>',
code_s14_agents: '<code>Always use the OpenAI developer documentation MCP server for OpenAI API, ChatGPT Apps SDK, Codex, and related product questions unless I explicitly say otherwise.</code>',
code_s14_prompt: '<code><span class="c"># 연결 후 테스트 프롬프트</span>\nOpenAI Responses API의 tools 요청 스키마를\nOpenAI developer docs에서 찾아서\n필수 필드만 bullet 5개로 요약해줘.</code>',
code_s14_scope: '<code><span class="c"># 전역 설정: CLI와 IDE Extension이 공유</span>\n~/.codex/config.toml\n\n[mcp_servers.openaiDeveloperDocs]\nurl = "https://developers.openai.com/mcp"\n\n<span class="c"># 프로젝트에만 적용하고 싶다면</span>\n./.codex/config.toml</code>',
code_s15_session: '<code>codex resume --last <span class="c"># 마지막 세션</span>\ncodex resume <span class="c"># 목록에서 선택</span>\ncodex resume <ID> <span class="c"># 특정 세션</span>\n\n<span class="c"># 세션 내 명령</span>\n/resume /fork /agent /status /compact\n/permissions /mcp /review /feedback</code>',
code_s16_exec: '<code><span class="c"># codex exec - 실행 후 자동 종료</span>\ncodex exec "Review this PR" --output stdout\ncodex exec "Fix all TODOs" --full-auto</code>',
code_s17_t1: '<code><span class="c"># ❌</span> 코드를 고쳐줘\n<span class="c"># ✅</span> @src/utils/date.ts 의 formatDate에서\nKST 처리 버그를 Intl.DateTimeFormat으로 수정해</code>',
code_s17_t2: '<code><span class="c"># 1단계:</span> User 모델과 타입 정의\n<span class="c"># 2단계:</span> CRUD API 엔드포인트\n<span class="c"># 3단계:</span> 인증 미들웨어</code>',
code_s17_t3: '<code>이 기능을 구현하고 단위 테스트도 작성해.\n린팅 통과 확인하고 기존 테스트도 돌려봐.</code>',
code_s17_t4: '<code><span class="c"># Codex 작업 중 방향 전환</span>\n> 잠깐, Zod 스키마를 대신 사용해줘\n<span class="c"># → 즉시 방향 전환</span></code>',
code_s18_adv: '<code><span class="c"># 모델 선택</span>\n/model gpt-5.5 <span class="c"># 최우선 추천 (보이면)</span>\n/model gpt-5.4 <span class="c"># 대체 선택지 / API key 경로</span>\n/model gpt-5.3-codex <span class="c"># cloud / review 비교용</span>\n\n<span class="c"># 플러그인 디렉터리 (CLI)</span>\n/plugins\n\n<span class="c"># 멀티 디렉토리</span>\ncodex --add-dir /path/to/libs\n\n<span class="c"># 웹 검색(인터랙티브에서 --search 전역 옵션 사용)</span>\ncodex --search "optimize with Next.js 15"\n\n<span class="c"># 컨텍스트 압축</span>\n/compact</code>',
},
en: {
nav_guide: "Guide",
nav_use_cases: "Use cases",
hero_badge_prefix: "Reviewed against official docs",
hero_title: "OpenAI Codex for<br/>Real-World Workflows",
hero_sub: "CLI · Desktop App (macOS/Windows) · IDE Extension · Web · Mobile · SDK automation<br/>Reorganized around the latest official docs so first-time and advanced users can choose the right surface, model, sandbox setup, MCP flow, and automation path faster.",
hero_cta: "Open the Fast Start Paths",
hero_cta_secondary: "View Practical Prompting Guide",
hero_jump_label: "Jump to a chapter",
hero_panel_kicker: "Launch paths",
hero_panel_title: "Start from the route you need right now",
hero_panel_desc: "Pick the most relevant path first, then drop into the sections below without having to decode the whole guide at once.",
hero_path_1_tag: "First run",
hero_path_1_title: "Install and finish a first task in 10 minutes",
hero_path_1_desc: "Move from app install and sign-in to one local task and a Git checkpoint.",
hero_path_2_tag: "Team setup",
hero_path_2_title: "Lock project rules and tool access early",
hero_path_2_desc: "Standardize AGENTS.md, config.toml, MCP, and approval policy before scaling usage.",
hero_path_3_tag: "Prompting",
hero_path_3_title: "Prompt contracts that survive longer tasks",
hero_path_3_desc: "Use section 17 for tool loops, verification, and runtime tuning patterns that hold up in real work.",
hero_signal_1: "chapters",
hero_signal_2: "live language toggle",
hero_signal_3: "official-doc review",
toc_title: "Guide Map",
toc_lead: "Start with the chapter map for orientation, then continue into section 17 when you need more prompting depth.",
toc_home_1_t: "Foundations",
toc_home_1_d: "What Codex is, how the product line fits together, and which models matter.",
toc_home_2_t: "Setup and Surfaces",
toc_home_2_d: "Installation, auth, and the practical differences between CLI, App, IDE, and Web.",
toc_home_3_t: "Execution Model",
toc_home_3_d: "Approval modes, slash commands, and how steering works during execution.",
toc_home_4_t: "Context and Tools",
toc_home_4_d: "Use AGENTS.md, config.toml, and MCP to shape behavior and tool access.",
toc_home_5_t: "Sessions and Automation",
toc_home_5_d: "Thread management, codex exec, web auto-review, and CI/CD workflows.",
toc_home_6_t: "Prompting Codex Agents",
toc_home_6_d: "Execution contracts, tool loops, runtime tuning, and reusable starter prompts in one embedded chapter.",
toc_home_7_t: "Decision Help",
toc_home_7_d: "Advanced usage, FAQ, and primary references.",
home_nav_kicker: "Guide map",
home_nav_title: "Codex 101 chapters",
home_nav_lead: "Move from reviewed facts to surface selection, execution rules, automation, and prompting in the order most teams actually apply them.",
overview_kicker: "Start here",
overview_title: "Reading paths for first-time and advanced users",
overview_lead: "This section starts with two things: which model/surface choices are current, and where first-time and professional users should begin. Today's edits are kept only in the Changelog near the bottom.",
overview_update_1_title: "Model and surface matrix rechecked",
overview_update_1_desc: "The June 3 run started by re-checking `codex/models` first, confirming that `gpt-5.5` is still the top recommendation, that `gpt-5.4` remains the fallback path, and clarifying how `gpt-5.4-mini`, `gpt-5.3-codex`, and `gpt-5.3-codex-spark` split fast local/subagent, Cloud/code-review, and Pro research-preview roles. The key API-key nuance remains: `codex/models` marks GPT-5.5 API Access as true, while the `codex/pricing` API Key table still lists it as not available.",
overview_update_2_title: "Pricing and credits refreshed",
overview_update_2_desc: "The guide keeps the `GPT-5.5` usage tables, `Free $0`, `Go $8`, `Plus $20`, `Pro from $100`, Plus web/CLI/IDE/iOS access, Business/Enterprise credits language, and the current Plus / Pro 5x / Pro 20x / Business / API Key pricing layout aligned. Pricing now lists Sites as available for Business/Enterprise and unavailable for Plus/Pro/API Key, while the Sites FAQ says it is free during preview. This run also keeps the split clear: API-key users should verify the actual Codex model picker and standard API pricing, while direct OpenAI API model access is documented separately.",
overview_update_3_title: "Quickstart, IDE, and mobile path tightened",
overview_update_3_desc: "The top path still keeps App-first onboarding separate from IDE/CLI + Docs MCP professional workflows, but the remote-connection story was updated against today's official docs. The current docs now say ChatGPT on iOS or Android can control an awake, online macOS or Windows Codex App host. Remote access uses the connected host's files, credentials, permissions, plugins, browser setup, Computer Use, and local tools, and setup may require the same account/workspace plus SSO, MFA, or passkeys. Business/Enterprise admins may need to enable Remote Control in workspace settings or grant it through RBAC. Windows hosts can be controlled from a phone or Mac, while the Windows Codex App still cannot control another computer.",
overview_update_4_title: "Enterprise announcements separated",
overview_update_4_desc: "The May 29 release notes add Windows Computer Use, Windows host remote control, responsiveness/browser stability, and Codex Profiles. I also rechecked the May 21 Enterprise/Edu release notes and May 22 Gartner announcement as enterprise context: Goal mode GA, Appshots, locked computer use, in-app browser annotations, admin Codex analytics, plugin sharing defaults, enterprise governance, sandboxing, RBAC, and auditability.",
overview_update_5_title: "Broader launch-post sweep",
overview_update_5_desc: "I re-read `developers.openai.com/codex`, `openai.com/index`, and `openai.com/codex` together. The June 2 `Codex for every role, tool, and workflow` launch is now reflected as official announcement language for role-specific plugins, Sites preview, and broader annotations, while current activation and operations notes are separated through the Sites and pricing docs. The May 29 Braintrust customer story and official `How OpenAI uses Codex` resource remain workflow patterns: customer requests to preview branches, Ask Mode, issue-shaped prompts, environment iteration, AGENTS.md, and Best-of-N. The May 27 Tax AI/Codex engineering post remains a self-improving agent operating pattern built from practitioner corrections, production traces, eval targets, bounded task environments, and human review.",
overview_update_6_title: "Docs MCP, API changelog, and config rechecked",
overview_update_6_desc: "Docs MCP, the config reference, and app/settings were rechecked around trusted project-scoped `.codex/config.toml`, the user-level requirement for machine-local provider/auth/profile/telemetry routing keys, shared CLI/IDE setup, the AGENTS.md steering snippet, `review_model`, `plan_mode_reasoning_effort`, granular approvals, `approvals_reviewer`, `default_permissions`, permissions profiles, sandboxed network proxy/requirements, keyring credential storage, forced login/workspace, `sqlite_home`, and hooks/plugin/skill controls. Settings > Profile remains documented as the place for lifetime tokens, peak tokens, streaks, longest task, and token activity. The OpenAI API changelog and deprecations docs were also checked: `gpt-5-codex`, `gpt-5.1-codex*`, and `gpt-5.2-codex` API snapshots are listed for July 23, 2026 shutdown with `gpt-5.5` as the substitute. Secure MCP Tunnel, workload identity federation, Admin API expansions, multiple IP allowlists, and Apps SDK MCP Apps lifecycle/server-instructions changes stay in the Platform/API/App developer lane rather than the Codex setup lane.",
overview_lens_beginner_kicker: "Beginner at a glance",
overview_lens_beginner_title: "Smallest path to a real first win",
overview_lens_beginner_desc: "Lock what “started” means before reading everything.",
overview_lens_beginner_s1: "Pick one surface first (app, IDE extension, or CLI) from Quickstart.",
overview_lens_beginner_s2: "Complete one local task with default sandbox and approvals.",
overview_lens_beginner_outcome: "Outcome: one successful task plus a safe Git checkpoint",
overview_lens_pro_kicker: "Professional at a glance",
overview_lens_pro_title: "Standardize team behavior first",
overview_lens_pro_desc: "Treat setup as project policy, not personal preference.",
overview_lens_pro_s1: "Define model, permissions, and done criteria in AGENTS.md and .codex/config.toml.",
overview_lens_pro_s2: "Connect Docs MCP by default so product/API answers stay grounded.",
overview_lens_pro_outcome: "Outcome: repeatable execution, review, and automation flows",
showcase_kicker: "Built with Codex",
showcase_title: "See real things built with Codex first",
showcase_desc: "OpenAI Developer Showcase collects apps, games, landing pages, and data visualization examples built with Codex. Use it to see what is possible and to get a feel for example prompts you can try right away.",
showcase_cta: "View Showcase",
overview_decision_kicker: "Decision first",
overview_decision_title: "A quick decision table for technical readers",
overview_decision_desc: "Before reading the full guide, pick the surface and model lane that match the work you are about to do.",
overview_decision_th_case: "Situation",
overview_decision_th_surface: "Recommended surface",
overview_decision_th_model: "Model lane",
overview_decision_th_next: "Read next",
overview_decision_local_case: "Small edits in your own project",
overview_decision_local_surface: "App Local or CLI",
overview_decision_local_model: "gpt-5.5, or gpt-5.4 if unavailable",
overview_decision_local_next: "05-06 setup and first run",
overview_decision_ide_case: "Continuous coding inside an IDE",
overview_decision_ide_surface: "IDE Extension",
overview_decision_ide_model: "Default recommended model + project context",
overview_decision_ide_next: "08 IDE Extension",
overview_decision_team_case: "Team rules and automation rollout",
overview_decision_team_surface: "CLI/App + AGENTS.md + MCP",
overview_decision_team_model: "gpt-5.5/gpt-5.4, mini for fast side tasks",
overview_decision_team_next: "12-16 operational setup",
overview_decision_review_case: "PR review and cloud tasks",
overview_decision_review_surface: "Web/Cloud",
overview_decision_review_model: "Cloud/review model lane",
overview_decision_review_next: "09 Web + 16 automation",
overview_beginner_tag: "First run",
overview_beginner_title: "From setup to the first task",
overview_beginner_desc: "Based on the Quickstart and App docs, this path focuses on project selection, Local mode, your first prompt, and Git checkpoints.",
overview_beginner_p1: "Start with sections 04-06 for install, sign-in, and the first CLI/App run.",
overview_beginner_p2: "Use section 10 to understand why sandboxing and approvals are separate controls.",
overview_beginner_p3: "Use section 14 to connect OpenAI Docs MCP and make product answers more reliable.",
overview_beginner_cta: "Go to setup",
overview_pro_tag: "Operations",
overview_pro_title: "From team rules to automation",
overview_pro_desc: "This path connects AGENTS.md, config.toml, session handling, non-interactive execution, and web review automation into a team workflow.",
overview_pro_p1: "Use sections 12-14 to lock behavior, configuration, and tool access at the project level.",
overview_pro_p2: "Use sections 15-16 to connect resume flows, codex exec, and PR review automation.",
overview_pro_p3: "Use section 17 to standardize output contracts, tool loops, and verification patterns.",
overview_pro_cta: "Go to operations",
overview_visual_kicker: "Visual guide",
overview_visual_title: "A beginner-friendly picture of how Codex fits together",
overview_visual_desc: "If this is your first time seeing Codex, scan this illustration first. It compresses the basic path into one image: choose a surface, start with a small task, stay safe with sandboxing and Git, then level up with AGENTS.md and Docs MCP.",
overview_visual_meta: "This guide image was created with the ImageGen skill.",
overview_visual_caption: "Beginner summary: choose a surface → start small → stay safe → grow with AGENTS.md and Docs MCP",
overview_plain_kicker: "Plain language",
overview_plain_title: "A simpler way to think about the jargon",
overview_plain_1: "`Local` means Codex works on your own computer, while `Cloud` means you send the job to a remote workspace.",
overview_plain_2: "`Sandbox` is the safety fence around what Codex can touch, and `Approval` is the extra prompt before it steps outside that fence.",
overview_plain_3: "`AGENTS.md` is the project playbook that says “please work this way in this repo.”",
overview_plain_4: "`MCP` is basically a connection cable that lets Codex reach docs, browsers, and external tools.",
overview_analogy_kicker: "Analogy",
overview_analogy_title: "It helps to think of Codex as a capable junior teammate",
overview_analogy_desc: "Instead of imagining a magic box that somehow writes everything perfectly, think of Codex as a strong coding assistant that can read files, make edits, run checks, and follow instructions. That is why the most natural beginner path is to start small, keep the safety rails on, and give it clearer project rules over time.",
overview_bridge_beginner_kicker: "Beginner terms map",
overview_bridge_beginner_title: "How to read this without confusion",
overview_bridge_beginner_1: "Local means on your machine, while Cloud means remote execution.",
overview_bridge_beginner_2: "Sandbox defines access boundaries, and Approval defines pause points.",
overview_bridge_beginner_3: "MCP is the integration layer that adds grounded docs/tools context.",
overview_bridge_pro_kicker: "Professional rollout focus",
overview_bridge_pro_title: "What matters most for teams",
overview_bridge_pro_1: "Lock trust boundaries first with sandbox_mode and writable_roots.",
overview_bridge_pro_2: "Control interruption frequency with approval_policy and rules.",
overview_bridge_pro_3: "Stabilize answer quality with AGENTS.md plus Docs MCP by default.",
overview_action_beginner_kicker: "10-minute starter plan",
overview_action_beginner_title: "Three checks right after setup",
overview_action_beginner_1: "Run codex and confirm you are in the correct project directory first.",
overview_action_beginner_2: "Keep default sandbox/approval settings and start with a small edit task.",
overview_action_beginner_3: "Create Git checkpoints before and after work so rollback stays easy.",
overview_action_beginner_cta: "Follow beginner checklist",
overview_action_pro_kicker: "Operational rollout plan",
overview_action_pro_title: "Three defaults to lock for teams",
overview_action_pro_1: "Declare role, guardrails, and done criteria explicitly in AGENTS.md.",
overview_action_pro_2: "Standardize model, reasoning effort, and approval policy via config.toml defaults.",
overview_action_pro_3: "Connect OpenAI Docs MCP so product/API answers stay grounded in docs.",
overview_action_pro_cta: "Follow operations checklist",
overview_fact_kicker: "Officially verified today",
overview_fact_title: "Today's updates",
overview_fact_date_prefix: "Updated on",
overview_fact_1_title: "Model recommendations aligned to surfaces",
overview_fact_1_desc: "Re-checking `codex/models` confirms that `gpt-5.5` is still the top recommendation. Start there when it appears in the Codex model picker, and use `gpt-5.4` when it has not rolled out to the account yet. One clarification now matters more in the guide: the Codex models page marks `gpt-5.5` API Access as true and mentions ChatGPT or API-key authentication, while the `codex/pricing` API Key usage table still lists `gpt-5.5` as not available. API-key users should verify the actual picker and standard API pricing before relying on it. `gpt-5.4-mini` stays the fast local/subagent option, Codex Cloud still belongs to the `gpt-5.3-codex` path, and `gpt-5.2` remains the main alternative model.",
overview_fact_2_title: "GPT-5.5 promoted, GPT-5.4 remains important",
overview_fact_2_desc: "The latest model story is now split more clearly. `gpt-5.5` is still the top recommendation for complex coding, computer use, knowledge work, and research workflows in Codex. In this refresh, the guide separates the Codex model picker, API-key authentication for Codex, and direct API model calls. The API itself exposes `gpt-5.5` and `gpt-5.5-pro` as separate models, while the API deprecations page lists `gpt-5-codex`-family snapshots for July 23, 2026 shutdown. API-key Codex users should compare the models page, pricing table, and actual product picker. `gpt-5.4` still matters for rollout, cost, or latency tradeoffs, while `gpt-5.4-mini` remains the efficient choice for lighter work and subagents.",
overview_fact_3_title: "Plans and team pricing clarified",
overview_fact_3_desc: "Quickstart still says every ChatGPT plan includes Codex. The current `codex/pricing` page keeps the Plus / Pro 5x / Pro 20x / Business / API Key tables and the published Plus local-message ranges (`gpt-5.5` 15-80, `gpt-5.4` 20-100, `gpt-5.4-mini` 60-350, `gpt-5.3-codex` 30-150). Plus now explicitly includes web, CLI, IDE extension, and iOS access. Sites is listed as a Business/Enterprise preview, and the pricing FAQ says it is free during preview. The important split is that API-key availability is described differently across `codex/models` and `codex/pricing`; direct OpenAI API access to `gpt-5.5` and `gpt-5.5-pro` belongs to the separate API Models/Pricing docs. The guide tells readers to check usage tables, token credit rates, the usage dashboard, App Settings > Profile lifetime-token and token-activity stats, the actual picker, and API model pricing without collapsing them into one bucket.",
overview_fact_4_title: "IDE and Windows paths clarified",
overview_fact_4_desc: "The IDE docs now cover both VS Code-style editors and JetBrains across macOS, Windows, and Linux. The Windows app docs now split the flow clearly: install from Microsoft Store or `winget install Codex -s msstore`, run the native agent in PowerShell, keep the native Windows sandbox on through Default permissions, restart after switching the agent to WSL2, and choose the integrated terminal separately. Read with the Windows sandbox engineering post, `elevated` is the preferred native sandbox and `unelevated` is the fallback when admin approval or policy blocks elevated setup.",
overview_fact_5_title: "App workflows reflected",
overview_fact_5_desc: "Cross-checking `Introducing the Codex app`, the April 16 `Codex for (almost) everything` post, and the current `app/features` / `app/automations` / `app/settings` docs helped move worktrees, built-in Git, the integrated terminal, skills, automations, thread automations, chats, the in-app browser, artifact preview, image generation, background computer use, memories, context-aware suggestions, broader plugins, role-specific plugins, broader annotations, multiple terminal tabs, SSH devboxes, and the summary pane earlier in the guide. Sites is now documented as a Business/Enterprise preview for saving and deploying OpenAI-hosted production URLs, with Enterprise RBAC/admin enablement, save-version versus deploy-version review, access modes, and runtime-secret management in the Sites panel called out. The June 1 OpenAI/AWS announcement and Codex Amazon Bedrock docs are reflected as an enterprise/provider deployment path with AWS-managed authentication, account controls, and billing, not as a ChatGPT plan-limit change.",
overview_fact_6_title: "Config/API/plugin flow refreshed",
overview_fact_6_desc: "The guide now keeps `review_model`, granular approvals, `approvals_reviewer`, top-level `web_search`, `tools.web_search`, `tools.view_image`, `tool_suggest`, `service_tier`, `personality`, `default_permissions`, Windows sandbox keys, `model_instructions_file`, `memories.disable_on_external_context`, `sqlite_home`, TUI settings, protected paths, JSON schema, app-connector controls, feature flags, permissions profiles, sandboxed network proxy/requirements, keyring credential storage, forced login/workspace controls, OpenTelemetry export, plugin packaging, Bedrock provider setup, and the split cloud runtime aligned with current docs. Project-scoped config loads only for trusted projects, and provider/auth/profile/telemetry routing keys belong in user-level config. The API changelog and deprecations docs belong to the Platform/API developer lane: OpenAI models in Amazon Bedrock use an OpenAI-compatible Responses API endpoint with model/feature support varying by AWS Region, while `gpt-5-codex`, `gpt-5.1-codex`, `gpt-5.1-codex-max`, and `gpt-5.2-codex` API snapshots are listed for July 23, 2026 shutdown with `gpt-5.5` as the substitute.",
overview_sources_label: "Source docs",
overview_source_1: "Codex Quickstart",
overview_source_16: "Codex Overview",
overview_source_2: "Codex Pricing",
overview_source_3: "Codex Best practices",
overview_source_4: "Codex Config Reference",
overview_source_5: "Agent approvals & security",
overview_source_6: "IDE extension",
overview_source_7: "Windows guide",
overview_source_20: "Docs MCP",
overview_source_8: "Codex Models",
overview_source_9: "GPT-5.5 announcement",
overview_source_14: "GPT-5.4 mini and nano announcement",
overview_source_10: "GPT-5.3-Codex-Spark announcement",
overview_source_13: "Codex App announcement",
overview_source_17: "Codex App Automations",
overview_source_18: "Codex App Features",
overview_source_19: "Codex Windows App",
overview_source_15: "Codex Plugins",
overview_source_21: "Codex Skills",
overview_source_11: "Codex for OSS",
overview_source_12: "AI-Native Engineering Team",
overview_source_22: "Codex for (almost) everything",
overview_source_23: "OpenAI API Models",
overview_source_24: "GPT-5.5 pro API Model",
overview_source_25: "Codex In-app Browser",
overview_source_26: "Pay-as-you-go pricing for teams",
overview_source_27: "Codex rate card",
overview_source_28: "Codex on Amazon Bedrock",
overview_source_29: "Advanced Account Security",
overview_source_30: "GPT-5.5 Instant",
overview_source_31: "Running Codex safely",
overview_source_32: "GPT-5.5-Cyber and Codex Security",
overview_source_33: "OpenAI API Changelog",
overview_source_34: "Codex Sites",
s1_refs: 'Sources: <a href="https://developers.openai.com/codex/" target="_blank">Codex overview</a> · <a href="https://developers.openai.com/codex/models/" target="_blank">Codex models</a> · <a href="https://developers.openai.com/api/docs/models" target="_blank">OpenAI API Models</a> · <a href="https://developers.openai.com/api/docs/changelog" target="_blank">OpenAI API changelog</a> · <a href="https://openai.com/index/introducing-gpt-5-5/" target="_blank">Introducing GPT-5.5</a> · <a href="https://openai.com/index/work-with-codex-from-anywhere/" target="_blank">Work with Codex from anywhere</a>',
s2_refs: 'Sources: <a href="https://developers.openai.com/codex/quickstart/" target="_blank">Quickstart</a> · <a href="https://developers.openai.com/codex/app" target="_blank">Codex app</a> · <a href="https://developers.openai.com/codex/ide" target="_blank">IDE extension</a> · <a href="https://developers.openai.com/codex/cli" target="_blank">Codex CLI</a>',
s3_refs: 'Sources: <a href="https://developers.openai.com/codex/models/" target="_blank">Codex models</a> · <a href="https://developers.openai.com/codex/pricing/" target="_blank">Codex pricing</a> · <a href="https://developers.openai.com/api/docs/models" target="_blank">OpenAI API Models</a> · <a href="https://developers.openai.com/api/docs/models/gpt-5.5-pro" target="_blank">GPT-5.5 pro API model</a>',
s4_refs: 'Sources: <a href="https://developers.openai.com/codex/quickstart/" target="_blank">Quickstart</a> · <a href="https://developers.openai.com/codex/pricing/" target="_blank">Codex pricing</a> · <a href="https://help.openai.com/en/articles/20001106-codex-rate-card" target="_blank">Codex rate card</a> · <a href="https://developers.openai.com/codex/speed" target="_blank">Speed</a> · <a href="https://openai.com/index/codex-flexible-pricing-for-teams/" target="_blank">Pay-as-you-go pricing for teams</a>',
s5_refs: 'Sources: <a href="https://developers.openai.com/codex/quickstart/" target="_blank">Quickstart</a> · <a href="https://developers.openai.com/codex/windows" target="_blank">Windows guide</a> · <a href="https://developers.openai.com/codex/app/windows" target="_blank">Windows app</a> · <a href="https://developers.openai.com/codex/ide" target="_blank">IDE extension</a> · <a href="https://openai.com/index/building-codex-windows-sandbox/" target="_blank">Windows sandbox engineering</a> · <a href="https://openai.com/index/running-codex-safely/" target="_blank">Running Codex safely</a>',
s6_refs: 'Sources: <a href="https://developers.openai.com/codex/quickstart/" target="_blank">Quickstart</a> · <a href="https://developers.openai.com/codex/cli" target="_blank">Codex CLI</a> · <a href="https://developers.openai.com/codex/cli/slash-commands/" target="_blank">CLI slash commands</a>',
s7_refs: 'Sources: <a href="https://developers.openai.com/codex/app" target="_blank">Codex app</a> · <a href="https://developers.openai.com/codex/app/features" target="_blank">App features</a> · <a href="https://developers.openai.com/codex/app/browser" target="_blank">In-app browser</a> · <a href="https://developers.openai.com/codex/app/automations" target="_blank">App automations</a> · <a href="https://openai.com/index/codex-for-almost-everything/" target="_blank">Codex for (almost) everything</a>',
s8_refs: 'Sources: <a href="https://developers.openai.com/codex/ide" target="_blank">IDE extension</a> · <a href="https://developers.openai.com/codex/ide/features" target="_blank">IDE features</a> · <a href="https://developers.openai.com/codex/ide/slash-commands" target="_blank">IDE slash commands</a> · <a href="https://developers.openai.com/codex/app/windows" target="_blank">Windows app</a>',
s9_refs: 'Sources: <a href="https://developers.openai.com/codex/quickstart/" target="_blank">Quickstart</a> · <a href="https://developers.openai.com/codex/cloud" target="_blank">Codex cloud</a> · <a href="https://developers.openai.com/codex/remote-connections" target="_blank">Remote connections</a> · <a href="https://developers.openai.com/codex/integrations/github" target="_blank">GitHub integration</a> · <a href="https://openai.com/index/work-with-codex-from-anywhere/" target="_blank">Work with Codex from anywhere</a> · <a href="https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes" target="_blank">Enterprise/Edu release notes</a>',
s10_refs: 'Sources: <a href="https://developers.openai.com/codex/agent-approvals-security" target="_blank">Agent approvals & security</a> · <a href="https://developers.openai.com/codex/config-reference/" target="_blank">Config reference</a> · <a href="https://developers.openai.com/codex/windows" target="_blank">Windows guide</a> · <a href="https://openai.com/index/building-codex-windows-sandbox/" target="_blank">Windows sandbox engineering</a>',
changelog_kicker: "Changelog",
changelog_title: "Today's Updates",
changelog_item_1: "Reconfirmed `gpt-5.5` as the top recommendation and separated the `codex/models` API Access flag from the `codex/pricing` API Key usage table.",
changelog_item_2: "Realigned Pro $100 2x promo timing, Pro $200 20x/25x Plus limit wording, Go/iOS plan wording, credit scope, and the warning that API-key users should verify the actual picker plus standard API model pricing.",
changelog_item_3: "Updated Remote connections for the current official wording: ChatGPT iOS/Android or a Mac can control a Windows host, while the Windows App still cannot control another computer.",
changelog_item_4: "Added the May 29 release-note updates for Windows Computer Use, Windows host remote control, Codex Profiles, and responsiveness/browser stability.",
changelog_item_5: "Added the June 2 role-specific plugins, Sites preview, and broader annotations announcement, while keeping Bedrock and Platform/API items in their separate lanes.",
changelog_item_6: "Chrome social verification confirmed official X/LinkedIn repeats of the announcement; Threads did not add a doc-backed claim, and unverified social claims were not added.",
overview_update_7_title: "Social tip check limited",
changelog_sources_toggle: "How sources are placed",
changelog_sources_note: "Instead of keeping one long source dump at the bottom, the guide now points you to the small `Sources` line under each section. The changelog should stay short, while the relevant section holds the links you are most likely to verify next.",
s1_title: "Codex at a Glance",
s1_intro: "OpenAI Codex is a coding agent that can read code, make changes, run checks, and explain the result.",
s1_philosophy: '"From writing code → finishing work"',
s1_desc: "Model, pricing, and setup details come later. This section only gives you the main surfaces and milestones.",
s1_timeline_title: "Key updates",
th_date: "Date", th_event: "Event",
tl1: "Codex CLI open-sourced",
tl2: "Cloud sandbox and Codex-1 introduced",
tl3: "Codex CLI Rust rewrite beta released",
tl4: "GPT-5 Codex GA announced at DevDay 2025",
tl5: "GPT-5.2-Codex released (multi-file optimization, context compaction)",
tl6: "IDE Extension expands Codex into editors",
tl7: "Codex macOS App released",
tl8: "GPT-5.3-Codex released",
tl9: "Codex-Spark research preview announced",
tl10: "Windows App expands Codex on desktop",
tl11: "GPT-5.4 introduced and promoted as Codex's recommended default model",
tl12: "GPT-5.5 promoted as Codex's top recommended model",
tl13: "Codex remote coordination preview launched in the ChatGPT mobile app",
s2_title: "Codex Product Suite at a Glance",
s2_cli: "Terminal-based (TUI) local agentic coding & CI/CD automation. Released Apr 2025.",
s2_app: "Desktop app for macOS and Windows. It acts as a command center for agents with built-in worktrees, an integrated terminal, diff review, plus skills, automations, and personality controls.",
s2_ide: "IDE agent across VS Code, Cursor, Windsurf, and JetBrains with both local work and cloud delegation.",
s2_cloud: "ChatGPT web/mobile surfaces. Web covers GitHub environments, cloud coding tasks, diff review, automatic PR/code review, and `@codex` delegation; the mobile preview lets you continue threads, approvals, model changes, and screenshot/terminal/diff/test review from a phone while Codex runs on a Mac or remote environment.",
s3_title: "Supported Models",
s3_models_title: "Current model guidance",
s3_rec_title: "Recommended Models",
s3_alt_title: "Alternative Models",
th_model: "Model", th_feature: "Features", th_use: "Use Case", th_note: "Note",
m0_f: "Newest frontier model for complex coding, computer use, knowledge work, and research workflows in Codex. In Codex, it is currently the top recommendation for ChatGPT-authenticated users.", m0_u: "Most important high-difficulty work",
m1_f: "The flagship fallback when GPT-5.5 is not available in the model picker yet, or when cost and latency matter more. It still combines strong coding, reasoning, and tool use in a balanced way, and it is also the first mainline model where built-in computer use was clearly introduced.", m1_u: "Professional development work and broader workflows",
m1mini_f: "Fast, efficient mini model for lighter coding tasks and subagents.",
m2_f: "Industry-leading coding model for complex software engineering. Its coding capabilities now also power GPT-5.4.", m2_u: "Complex real-world SW engineering",
m3_f: "Text-only research preview model optimized for near-instant, real-time coding iteration. Available to ChatGPT Pro users.", m3_u: "Real-time coding, fast iteration",
s3_rec_note: "💡 The latest `codex/models` guidance says to start with `gpt-5.5` when it appears in your picker. Today that page marks GPT-5.5 API Access as true, while the `codex/pricing` API Key table still lists it as not available, so API-key users should verify the actual Codex picker and standard API pricing. If cost/latency matters more, keep using `gpt-5.4`. In the API, `gpt-5.5` and `gpt-5.5-pro` are available, with Pro aimed at higher-accuracy Responses API and Batch API work. Use `gpt-5.4-mini` for faster lighter tasks or subagents, `gpt-5.3-codex` for complex software engineering and current Codex Cloud work, and `gpt-5.3-codex-spark` for near-instant iteration. The main current alternative on that page is `gpt-5.2`.",
computer_use_kicker: "Computer use",
computer_use_title: "Treat computer use as its own feature chapter",
computer_use_desc: "Per the official docs, `gpt-5.4` is the first mainline model with built-in computer-use capabilities. In practice, that means the model got better at a loop where it sees the screen, chooses an action, executes it through a harness, then checks the updated screen again.",
computer_use_p1: "This matters most for automation that needs to work directly through a UI instead of only replying in text.",
computer_use_p2: "The core loop is `screenshot → action → screenshot → verify/fix`.",
computer_use_p3: "It is especially useful for browser testing, form flows, and multi-step UI tasks that are easier to demonstrate visually.",
computer_use_p4: "If `Plugins` are the toolbox that connects outside services, `computer use` is the feature that directly works with what is visible on screen.",
computer_use_p5: "If `Skills` explain how a workflow should be done, `computer use` is one of the execution methods that can actually click, type, and navigate the UI.",
computer_use_cta: "Open the official Computer use guide",
computer_use_caption: "English uses OpenAI’s official diagram; Korean uses a localized explainer visual.",
m4_f: "Previous general-purpose model for coding and agentic tasks across industries and domains. It now sits behind the `gpt-5.5` / `gpt-5.4` default pair as an alternative option.", m4_u: "Cross-industry/domain",
s3_other_note: "⚠️ Codex works best with the models listed above. You can also use any model/provider supporting Chat Completions or Responses APIs, but Chat Completions API support will be removed in future releases.",
model_show: "Show details", model_hide: "Hide details",
feat_cap: "Capability", feat_spd: "Speed",
feat_cli: "Codex CLI & SDK", feat_app: "Codex app & IDE extension",
feat_cloud: "Codex Cloud", feat_credits: "ChatGPT Credits", feat_api: "API Access",
s3_cfg_title: "🛠️ Model Configuration Tip",
s3_cfg_default: "You can set your default model in config.toml. If not specified, a recommended model is used by default.",
s3_cfg_temp: "Use <code>/model</code> command or <code>codex -m model-name</code> flag to change models temporarily in CLI. In the IDE extension, use the model selector below the input box.",
s3_cfg_cloud: "Currently, you can't change the default model for Codex cloud tasks.",
s4_title: "System Requirements & Pricing",
s4_req_title: "System Requirements", s4_price_title: "Main plans and access paths",
s4_git: "Strongly recommended", s4_acct_label: "Account", s4_custom: "Custom pricing",
s4_note_access: "The official docs still say every ChatGPT plan includes Codex. For first-time users, the easiest reading order is `Free $0` -> `Go $8` -> `Plus $20` -> `Pro from $100`; for teams, add the Business/Enterprise Codex-only pay-as-you-go seat path. The Pricing page breaks usage out by Plus / Pro 5x / Pro 20x / Business / API Key and still shows model-specific five-hour limits. On Plus, the published local-message ranges are `gpt-5.5` 15-80, `gpt-5.4` 20-100, `gpt-5.4-mini` 60-350, and `gpt-5.3-codex` 30-150. The key distinction is that API-key availability is described differently across `codex/models` and `codex/pricing`; direct OpenAI API access to `gpt-5.5` and `gpt-5.5-pro` is covered by the separate API Models/Pricing docs. In practice, compare message/task limits, token credit rates, the usage dashboard, the actual picker, and API model pricing without mixing those scopes. Built-in image generation uses the same general limits about 3-5x faster on average, and Fast mode consumes credits faster on supported models.",
th_plan: "Plan", th_price: "Price", th_codex: "Codex scope",
s4_codex_free: "Explore Codex on quick coding tasks",
s4_codex_go: "Use Codex for lightweight coding tasks",
s4_codex_plus: "Web/CLI/IDE/iOS + latest models (including GPT-5.5) + cloud review/Slack + token-based credits",
s4_codex_pro: "Spark preview + priority processing + Pro 5x / Pro 20x tables; May 31 promo boosts need dashboard verification",
s4_codex_business: "Standard Business seat or Codex-only pay-as-you-go seat + larger VMs + credits + SAML SSO/MFA",
s4_codex_enterprise: "Business features + Codex-only pay-as-you-go path + priority processing + RBAC/audit/governance",
s4_usage_based: "Token-usage based",
s4_codex_api: "CLI/IDE/SDK only, no cloud features. With API-key auth, verify the actual picker and standard API pricing",
s5_title: "Installation & Authentication",
s5_cli_install: "CLI Installation", s5_auth: "Authentication",
s5_app_install: "Desktop App Installation (macOS/Windows)",
s5_app1: "macOS: Download from openai.com/codex (.dmg). If you're using an Intel-based Mac, choose the Intel build.",
s5_app2: "Windows: Install from the Microsoft Store or run `winget install Codex -s msstore`",
s5_app3: "Launch the app and sign in with a ChatGPT account or API key (API-key sign-in can limit cloud-thread and some credits-based features). For high-risk accounts or sensitive projects, also review Advanced Account Security in ChatGPT web security settings.",