fix: Skill 未引导正确处理环境ID的简写与完整形式

[binggg]

Attribution issue

• issueId: issue_mnovxo2g_pc6qna
• category: skill
• canonicalTitle: Skill 未引导正确处理环境ID的简写与完整形式
• representativeRun: application-js-react-cloudbase-phone-signup/2026-04-09T21-19-12-8cfnsp

Automation summary

• root_cause: This is a real product defect in the skill guidance. The canonical CloudBase guidance told agents to use the environment ID from envQuery, but it did not explicitly instruct them to resolve alias-like shorthand or nicknames to the canonical full EnvId before calling auth.set_env, initializing SDKs, generating config, or building console links. That gap can produce user-visible mistakes in normal usage, not just in evaluation.
• changes: Added minimal product-facing guidance in the main CloudBase guideline, the platform overview skill, and the editor compatibility guide to require envQuery(action=list, alias=..., aliasExact=true) when the conversation only contains an environment alias or shorthand, and to stop for clarification when the alias is ambiguous. Also added a focused regression test that asserts this guidance is present in those canonical surfaces.
• validation: Reviewed the diff to confirm it adds guidance only, with no benchmark-only aliases or duplicate field variants. Ran a small static validation script to verify the new alias-resolution instructions and regression test text are present; it passed (static-validation: ok). I did not run the Vit

Changed files

• config/source/editor-config/guides/cloudbase-rules.mdc
• config/source/guideline/cloudbase/SKILL.md
• config/source/skills/cloudbase-platform/SKILL.md
• tests/skill-quality-standards.test.js

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[binggg]

Attribution post-PR evaluation

• visibility: internal identifiers, run ids, and private links are intentionally omitted
• attempt: 5
• eval_scope: primary_plus_regression
• overall: FAILED
• summary: at least one planned evaluation case failed
• updated_at: 2026-04-21T02:34:52.210Z

Cases

• [FAILED] — primary — evaluation failed
• [PENDING] — regression 1 — queued

[binggg]

Superseded by #579, which is a cleaner rebase on latest main (drops .claude mirror diffs, restores an accidentally removed section in cloudbase-rules.md, and squashes 7 retry commits into 1). Closing this PR.

New PR has been validated by a targeted evaluation run: env-destroy case score 0.401 → 0.833 with agent correctly resolving the alias via envQuery before DestroyEnv.