Skip to content

Phase-7-Carousel-Intel.md

Latest commit

 

History

History
162 lines (132 loc) · 7.07 KB

Phase-7-Carousel-Intel.md

File metadata and controls

162 lines (132 loc) · 7.07 KB
Raw

Phase 7 — Carousel menu + Intel + Recon

Phase 7 overhauls the front of the TUI (Overworld), adds a 9-layer OSINT aggregator screen (Intel), and ships a 7-tab action console for on-demand OSINT lookups (Recon). M1–M7 of the internal plan landed together; M8 polished + documented.

Overworld (front door)

ScreenId::Overworld is now at index 0 of ScreenId::ALL so the Tab cycle always opens here (the "front door" metaphor from Bruce firmware). The screen is a single-pane carousel: every visible sidebar entry (Editor excluded) is rendered as a tile in a width-bucketed grid:

Width Cols
≤ 80 2
≤ 120 3
≤ 160 4
> 160 5

Keys:

  • Tab / arrow keys — move the cursor between tiles
  • Enter — switch to the focused screen (mirrors switch_screen with app.current + wm::WindowKind)
  • Esc — show an info toast (never quits)
  • Esc-on-Overworld — do not propagate

Tab-strip preview indicator (M3)

app.tab_cursor already owned cycle_tab_cursor / commit_tab_cursor / clear_tab_cursor. M3 wired those into the tab-strip area: ui::chunks() returns Option<Rect> for the strip (collapses to None when area.height < 10); main::draw() paints the strip when Some and writes app.tab_strip_rect. Key handling:

  • Tab / BackTab on the content side → cycle_tab_cursor(forward)
  • Enter → commit_tab_cursor() (jumps to the highlighted tab)
  • Esc on the content side → clear_tab_cursor() (drops the highlight, returns to current)

Intel screen (M4 + M5)

ScreenId::Intel (18th entry). Two-pane layout (Horizontal: Percentage(28) left grid | Min(40) right detail). Left = 9 OSINT layer rows (Flights / Earthquakes / Fires / Weather / Satellites / News / CCTV / Maritime / Conflicts) plus a sentinel chip in the title bar. Right = selected layer's summary line, sentinel chip, entity count, "last ok" timestamp, and the head of the upstream JSON.

Data path (M5)

cyberdeck_intel::refiller::spawn_all() runs one tokio::spawn task per LayerId, each at its layer's staggered poll interval (30 s – 21 600 s so consecutive layers never share an interval). Every fetch produces a Snapshot { layer, status, sentinel, summary, entity_count, raw } which is pushed through an mpsc → wrapped as Action::IntelSnapshot → handled in main.rs → App::intel_snapshots.insert(layer, snap). The render path reads from intel_snapshots first and falls back to the M4 fixture for any layer not yet populated.

The footer line reads intel: N/M layers live · K <SENTINEL> where N = layers with Ok status and the worst sentinel is rolled up via cyberdeck_intel::worst_sentinel.

Recon screen (M7)

ScreenId::Recon (19th entry). Single-pane — the output IS the screen, so has_right_pane() returns false. Seven tabs in stable order:

Tab Glyph Primitive Endpoint
DNS D dig +short system dig
WHOIS W whois system whois
IP I ureq::get ip-api.com ip-api.com/json
SSL S openssl s_client -brief system openssl
CVE C bundled NVD fixture local CSV
CRYPTO ₿ bundled risk table local CSV
SANCTIONS ⚖ bundled OFAC SDN mirror local CSV

SSRF gate

Every primitive that resolves a user-supplied target to a network endpoint runs through cyberdeck_intel::recon::ssrf::check_ip before any process spawn or HTTP call. Reject table:

Range Rule
127.0.0.0/8 loopback
10.0.0.0/8 RFC1918 private
172.16.0.0/12 RFC1918 private
192.168.0.0/16 RFC1918 private
169.254.0.0/16 link-local
0.0.0.0/8 "this network"
224.0.0.0/4 multicast
240.0.0.0/4 reserved
::1 IPv6 loopback
:: IPv6 unspecified
fe80::/10 IPv6 link-local
fc00::/7 IPv6 ULA
ff00::/8 IPv6 multicast

The handler emits a SsrfError::Blocked { addr, rule }. Six property tests (proptest, 256 cases each) pin the reject bands in CI.

Keymap

  • Tab / BackTab — cycle tabs
  • printable chars — append to query buffer (cap 256)
  • Enter — runs the active arm
  • Esc — clears the query + output
  • j / k (and arrow keys) — scroll the output area

Output buffer caps at 4 KiB so a 1 MiB WHOIS response can't pin the renderer; truncations are signalled with (… output truncated) at the head of the buffer.

CLI parity

cyberdeck recon <arm> <query> mirrors the screen's seven tabs: dns, whois, ip, ssl, cve, crypto, sanctions. Errors surface as {"ok":false, "error":{"message":"…"}} with the SSRF rule tag intact (e.g. "refused to target 127.0.0.1: loopback (127.0.0.0/8)").

Known issues

  • Six pre-existing bin tests (bluetooth_passkey_rejects_letters, esc_in_editor_closes_editor, esc_in_files_goes_up_a_folder, esc_in_logs_clears_active_filter, keymap_capture_rejects_conflict, choice_modal_cursor_wraps_and_enter_dispatches) predate Phase 7; they are not regressions from this work.
  • The refiller starts immediately on App::new. If you launch the TUI without internet, layers will surface as Error { reason: "fetch: …" } rows after the first poll. This is intentional — the Intel screen falls back to the M4 fixture for any layer not yet populated, so first-paint still looks correct.
  • cyberdeck recon ip <hostname> goes through ip-api.com, which rate-limits anonymous calls. Production deployments behind a proxy should set HTTPS_PROXY and tune polling accordingly.

Files added or substantively touched

  • crates/intel/{lib,refiller}.rs + 11 layer/recon modules
  • crates/intel/testdata/{sanctions_sample.csv,crypto_risk.csv}
  • crates/tui/src/screens/{overworld,intel,recon}/ (new)
  • crates/tui/src/screens/mod.rs + crates/tui/src/main.rs
  • crates/tui/src/app/{screen,action}.rs + crates/tui/src/app.rs
  • crates/tui/src/ui/{mod,tab_strip}.rs
  • crates/cli/src/commands/{intel,recon}.rs + commands/mod.rs
    • lib.rs
  • crates/cli/tests/cli_dispatch.rs
  • crates/daemon/src/{rpc,handlers}.rs
  • crates/daemon/Cargo.toml (added cyberdeck-intel dep)
  • crates/cli/Cargo.toml (added cyberdeck-intel dep)
  • crates/intel/Cargo.toml (added ureq, csv, proptest)