菜单
Audit the project against the R.E.S.T. framework (TRAE) — Reliability, Efficiency, Security, Traceability. Reads ledger entries, hook state, and artifact presence across long-session + policy-gateway + context-engine + traces. Single PASS/WARN/FAIL table. Cross-cut meta-surface.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when the user describes a multi-step feature or refactor and you need to decide whether to handle it solo, dispatch a parallel `/fan-out`, or run a `/worktree-team` planner→generator→reviewer pipeline. Outputs a routing recommendation with the reasoning behind it.
Use whenever the user asks to apply the same operation to many independent files or components — bulk migration, batch refactor across controllers, parallel exploration of unrelated subsystems. Dispatches one subagent per file with a shared prompt template, then collects the results.
Use to toggle the safe-mode flag at .claude/safe-mode. The flag is written automatically by consecutive-failure-guard once tool failures hit FORGE_SAFE_MODE_THRESHOLD (default 5), and block-destructive.sh denies every Bash/Write/Edit while it exists. /safe-mode off clears the flag, logs the exit, and prompts /postmortem; /safe-mode on enters manually; /safe-mode status reports current state.
Run the 5-stage Token Transformation Pipeline (TRAE §5.2.2) over the current session state — Collection, Ranking, Compression, Budgeting, Assembly. Emits a concrete next-action recommendation (/compact, /lean-md, or /progress-log + fresh session) instead of a generic warning.
Audit or scaffold a CLAUDE.md against the Karpathy 4-section structure (Think Before Coding · Simplicity First · Surgical Changes · Goal-Driven Execution). Companion to /lean-md — structure audit, not size trim.
Print every policy enforcement point declared in plugins/diagnostics/registry/policies.json — id, verdict, plugin, hook event, severity, bypass — grouped by verdict. Single discoverable index of what the harness blocks, anchors, nudges, or logs at runtime.
| name | rest-audit |
| description | Audit the project against the R.E.S.T. framework (TRAE) — Reliability, Efficiency, Security, Traceability. Reads ledger entries, hook state, and artifact presence across long-session + policy-gateway + context-engine + traces. Single PASS/WARN/FAIL table. Cross-cut meta-surface. |
| when_to_use | Periodic checkup, before releases, or when investigating a session regression that may span multiple plugins (a missed gate, a leaked secret, a token blowout). Do NOT use for structural drift — that's `/entropy-scan`; this skill reports outcomes (R/E/S/T axes), not file-level mismatches. |
| disable-model-invocation | true |
| allowed-tools | ["Read","Grep","Glob","Bash"] |
| logical | report shows PASS / WARN / FAIL for each of the 4 R.E.S.T. axes with worst-axis as overall verdict |
Outcome-oriented audit surface. Unlike /entropy-scan (structural — drift/registration/invariants), this one reports against the four production-readiness outcomes from TRAE's R.E.S.T. framework: Reliability · Efficiency · Security · Traceability.
Every axis pulls from at least two existing plugins so a failure has a concrete remediation path.
For each axis, collect signals → classify PASS / WARN / FAIL → report one actionable next step.
| Check | PASS | WARN | FAIL |
|---|---|---|---|
claude-progress.txt exists AND has ≥1 entry | yes | 0 entries | missing |
init.sh exists AND executable | yes | non-exec | missing |
Ledger has ≥0 safe-mode-enter events in last 7 days | 0–1 | 2–3 | ≥4 (runaway) |
/postmortem coverage: ratio of safe-mode-exit ledger events followed by a memory entry within 24h | ≥0.8 | 0.5–0.8 | <0.5 |
consecutive-failure-guard present in hooks.json | present | — | missing |
Sources: .claude/lineage/ledger.jsonl, claude-progress.txt, init.sh, plugins/context-engine/hooks/hooks.json.
| Check | PASS | WARN | FAIL |
|---|---|---|---|
| CLAUDE.md size (root) | <150 lines | 150–300 | >300 |
/token-pipeline nudge surfaced last session | yes (hook log or user run) | no data | — |
| Memory topic count | <30 | 30–60 | >60 |
| Features.json pending/done ratio (if present) | <1.0 | 1.0–3.0 | >3.0 (backlog blowout) |
| Traces hook wiring present | present | — | missing |
Sources: CLAUDE.md, .claude/memory/MEMORY.md, .claude/features.json, plugins/traces/hooks/hooks.json.
| Check | PASS | WARN | FAIL |
|---|---|---|---|
policy-gateway plugin present | yes | — | missing |
block-destructive.sh + scan-secrets.sh both PreToolUse | both | one | neither |
rules.d/secrets.txt exists AND non-empty | yes | empty | missing |
Ledger policy-block events in last 30 days (sign the plugin is armed AND being used) | ≥1 | 0 but plugin present | plugin missing |
Sensitive files not in working tree (grep matches via /policy-audit pass 2) | 0 matches | 1–2 | ≥3 |
Sources: .claude-plugin/marketplace.json, plugins/policy-gateway/, ledger.
| Check | PASS | WARN | FAIL |
|---|---|---|---|
traces/hooks/hooks.json registers ≥3 collectors | ≥3 | 1–2 | 0 |
| Ledger entries in last 7 days | ≥5 | 1–4 | 0 |
Handoff-equivalent coverage: claude-progress.txt entries vs session count (sessions count = .claude/traces/session-*.jsonl file count or similar) | ≥0.7 | 0.3–0.7 | <0.3 |
/lineage-audit runs clean (invoke it as part of rest-audit; propagate failures) | clean | warnings | fails |
Sources: plugins/traces/hooks/hooks.json, .claude/lineage/ledger.jsonl, .claude/traces/ (if present), call /lineage-audit.
R.E.S.T. AUDIT — <UTC>
=============================================
Reliability [PASS/WARN/FAIL]
<check> [status] <detail>
Next: <one concrete action>
Efficiency [PASS/WARN/FAIL]
<check> [status] <detail>
Next: <one concrete action>
Security [PASS/WARN/FAIL]
<check> [status] <detail>
Next: <one concrete action> (e.g. "Run /policy-audit for deep dive")
Traceability [PASS/WARN/FAIL]
<check> [status] <detail>
Next: <one concrete action> (e.g. "Run /lineage-audit")
Overall: PASS / WARN / FAIL
Axis status = worst individual check. Overall = worst axis.
Input: ledger has 12 entries in the last 7 days (no safe-mode runaway), CLAUDE.md is 142 lines, block-destructive.sh + scan-secrets.sh both registered, 4 traces collectors registered, /lineage-audit runs clean.
Output:
R.E.S.T. AUDIT — 2026-05-19T10:00Z
=============================================
Reliability PASS
claude-progress.txt: 14 entries [PASS]
safe-mode events last 7d: 0 [PASS]
Next: none
Efficiency PASS
CLAUDE.md size: 142 lines [PASS]
Memory topics: 18 [PASS]
Next: none
Security PASS
policy-gateway present [PASS]
block-destructive + scan-secrets [PASS both]
policy-block events last 30d: 3 [PASS]
Next: none
Traceability PASS
traces collectors: 4 [PASS]
/lineage-audit: clean [PASS]
Next: none
Overall: PASS
Input: policy-gateway present and armed, but no policy-block ledger entries in 30 days (plugin armed but unexercised); everything else green.
Output:
R.E.S.T. AUDIT — 2026-05-19T10:00Z
=============================================
Reliability PASS
Efficiency PASS
Security WARN
policy-gateway present [PASS]
block-destructive + scan-secrets [PASS both]
policy-block events last 30d: 0 [WARN — armed, never fired]
Next: Run /policy-audit to confirm scanners are reaching the working tree (a 0-count could mean clean repo OR misconfigured matchers).
Traceability PASS
Overall: WARN
Reads from everywhere — this is the meta-surface:
.claude/lineage/ledger.jsonl — safe-mode, policy-block, progress-log, SEPL eventsclaude-progress.txt — long-session coverage.claude/spec.md, .claude/features.json — living spec / testable reqs.claude/memory/MEMORY.md — memory bloat signalCLAUDE.md — efficiency signalplugins/*/hooks/hooks.json — wiring-presence signalsplugins/policy-gateway/rules.d/*.txt — armed-rules signalDelegates to:
/policy-audit for Security deep dive/lineage-audit for Traceability deep dive/token-pipeline for Efficiency remediation planning/postmortem for Reliability remediation (after safe-mode exits)Invoked by:
/entropy-scan includes this skill as a sub-check/evolve reads the FAIL list to seed proposals (e.g. if Security axis fails, propose adding a rule to rules.d/secrets.txt)/lineage-audit not installed → Traceability axis still runs (skips only the sub-call)./postmortem memory check requires reading memory index) — degrade to best-effort.