audit

A read-only report of recent automatic maintenance activity in a coding project. It reviews recorded actions, pending proposals, failed work, and the health of maintenance summaries.

In plain words
What is it for?
Use it to review the last 14, 7, or 30 days, find unfinished or unverified maintenance proposals, spot low-success skills, and check whether activity summaries are up to date.
Why use it?
It shows what automatic maintenance changed and whether anything needs attention, without changing the project itself.

Skill for Claude CodeCodex

Part of the core plugin — 37 skills, 4 agents, 11 hooks shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/ryanzhao1011/workframe/audit
Any agent
npx skills add ryanzhao1011/workframe --skill audit
Clone the repo
git clone --depth 1 https://github.com/ryanzhao1011/workframe

Made for: Claude Code, Codex.

Or install core, the plugin that ships this one along with the rest of its 37 skills, 4 agents, 11 hooks.

Per session 26 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,507 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00026 $0.01507
Opus 5 $0.00013 $0.00754
Sonnet 5 $0.00005 $0.00301
Haiku 4.5 $0.00003 $0.00151

Measured 2d ago against content hash d3511a49eee2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

audit scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Reads agent configuration directorieslowAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

python "$(cat .claude/workframe-state/plugin-root.txt)/bin/workframe-audit-board-drift"

Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.

plugins/core/skills/audit/SKILL.md · 113 lines

How it starts

The opening of the file, as written. The whole thing — 113 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/core:audit 审计维护活动

用途

用户显式 /core:audit 时执行。Claude 不会自动调用disable-model-invocation: true)。

回答五类问题:

  1. 最近 N 天自动发生了什么维护动作?(events.jsonl 聚合)
  2. 有没有 T3/T4 待处理项?(activity-state.pending_maintenance + proposals/pending/)
  3. 有没有失败的提案或低成功率的 skill?(skill-metrics.yaml)
  4. board summary 是否健康?(current drift + 近 30 天修复历史)
  5. skill-metrics 是否刷新?(skill_metrics_recomputed 事件 + 当前 generated_at

输入

参数格式(自由文本):

  • 无参数 → 默认近 14 天
  • 7d / 30d → 指定窗口

audit 保持纯只读——dismiss 等写操作迁到 /core:maintenance-review --dismiss <PM-ID>。这样 audit 与 maintenance-review 的边界清晰:audit 报告状态,maintenance-review 修改状态。

执行步骤

  1. Read .claude/workframe-state/activity-state.json — 拿 session_counter / dormant / pending_maintenance / recent_drift_repairs
  2. Read .claude/workframe-state/skill-metrics.yaml — 最近窗口的 skill/rule 汇总
  3. 读取 .claude/workframe-state/events.jsonl 近 N 天 — 按 type 分组统计(含 summary_drift_repaired / summary_drift_repair_skipped)
  4. Glob projects/proposals/pending/*.yaml — 列出待审批提案
  5. Glob projects/proposals/applied/*.yaml 过滤 verified: null — 列出待闭环验证
  6. Read .claude/workframe-state/session-digest-latest.md — 上次会话摘要
  7. Board summary drift check只读不改):
    • 调用只读检查命令(插件根路径从 plugin-root.txt 取,不依赖 PATH):
      python "$(cat .claude/workframe-state/plugin-root.txt)/bin/workframe-audit-board-drift"
      
    • 输出文本格式:
      actual: total=N, counts={pending:..,...}, unknown=[..]
      summary: total=M, counts={pending:..,...}
      drift: <empty | field1=summary→actual, field2=...>
      
    • audit 只解析 drift: 行决定是否展示 drift 状态;不调用 workframe-recompute-board-summary(那个会写文件,违反 audit 只读约束)。修复路径走 SessionStart drift check 自动 / workframe-recompute-board-summary 兜底命令
    • 调用方式说明:${CLAUDE_PLUGIN_ROOT} 在 agent Bash 上下文不可用,CC 官方的 plugin bin/ PATH 注入在部分环境也不生效(Windows + directory 订阅实测未注入),故统一走 plugin-root.txt 配方——该文件由 SessionStart hook 每次会话刷新为当前插件根(正斜杠绝对路径)。若环境里 workframe-audit-board-drift 恰好已在 PATH,裸调等价

输出格式

## 🔍 维护审计报告(近 <N> 天)

### 📊 事件统计
- skill_used: X 次(top: librarian ×5 / task-management ×3)
- rule_triggered: N/A(无 deterministic producer,字段保留仅为兼容;见 event-schema.json)
- user_correction: Z 次(problem 信号)
- task_blocked: W 次
- skill_metrics_recomputed: M 次(最近 generated_at: <ISO>)
- proposal_applied: A 次,proposal_verified: B 次(signal_met=true),proposal_failed: C 次(best_effort;依赖 self-iteration 执行)

### 🗂️ 提案状态
- pending 待审批:N 条
  - [PROP-...] confidence=0.75 | L2 | pattern=...
- applied 待验证:M 条(verify_by 到期前不闭合)

### 🏁 活动指标
- session_counter: xxx
- active_sessions_30: yy
- dormant: false (profile=normal)
- wake_up_pending: false
- last_digest_at: <ISO>

### ⚠️ pending_maintenance(按 severity 排序)
- 总数:N 条(open: N1 / closed: N2)
- critical: ...
- warn: ...
  - [PM-20260424-003] cadence_timeout — 距上次自迭代 10 天
- info: ...

> 关闭指定条目请使用 `/core:maintenance-review --dismiss <PM-ID>`(audit 保持只读,写操作归 maintenance-review)。

### 📐 Board Summary Drift 健康度

- **当前状态**:✅ summary 与 tasks 实际计数一致 / ⚠️ 发现 N 项 drift(current snapshot)
- **近 30 天修复历史**(来自 activity-state.recent_drift_repairs + events.jsonl):
  - 自动修复(summary_drift_repaired):X 次
  - 跳过修复(summary_drift_repair_skipped):Y 次
    - unknown_statuses_in_tasks: <list>
    - summary_block_not_found / recompute_failed / read_failed / parse_failed: <count>

如近 30 天 ≥3 次自动修复 → 提示:
> ⚠️ 频繁触发 drift 修复可能是 SessionEnd hook 经常未执行或超时。
> 大项目可设环境变量 `CLAUDE_CODE_SESSIONEND_HOOKS_TIMEOUT_MS=5000` 缓解。

如出现 unknown_statuses_in_tasks → 提示:
> ⚠️ 检查 `projects/board.yaml` 中 task status 拼写:<具体值>
> SessionStart drift check 不会自动修复含未知 status 的 board.yaml(C+ 第一保护条件)。

### ⚠️ 其他注意项
- 低成功率 skill: <list>

Read the full file on GitHub · 113 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 113 lines · 26 tokens per session scan A d3511a49eee2

Subscribe to this mod's changes

audit is a skill published in the GitHub repository ryanzhao1011/workframe (4 stars, last pushed 15d ago), licensed MIT. It adds 26 tokens to every session and 1,507 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

check-understanding

Phase quiz for AI Engineering from Scratch. Trigger with "quiz me", "test phase", "check my understanding", "do I know phase 3", or /check-understanding .

fancyboi999/ai-engineering-from-scratch-zh · 44 tokens

find-your-level

Interactive quiz that maps your AI/ML knowledge to a starting point in the 260-lesson, 20-phase AI Engineering from Scratch curriculum. Trigger phrases: "where should I start", "find my level", "what do I know", "which phase", "assess my knowledge", "placement test", "skip ahead".

fancyboi999/ai-engineering-from-scratch-zh · 71 tokens

audit-agents-skills

Audit Claude Code agents, skills, and commands for quality and production readiness. Use when evaluating skill quality, checking production readiness scores, or comparing agents against best-practice templates.

FlorianBruniaux/claude-code-plugins · 41 tokens

eval-agents

Audit Claude Code agents defined in .claude/agents/ for description specificity, model tier appropriateness, tools scoping, and system prompt quality. Detects dispatch ambiguity between agents, flags over-permissive tool grants, and checks for human-in-the-loop patterns that break programmatic orchestration. Use when…

FlorianBruniaux/claude-code-plugins · 93 tokens

eval-hooks

Audit Claude Code hooks defined in settings.json files for validity, performance safety, and correctness. Resolves each command against the filesystem, checks exit-code strategy for blocking hooks, flags missing timeouts, and reviews interactive vs async patterns. Use when setting up hooks for the first time…

FlorianBruniaux/claude-code-plugins · 78 tokens

check-cache-bugs

Audit Claude Code setup for cache bugs (CC#40524): sentinel, --resume/--continue, attribution header + ArkNill B3/B4/B5.

FlorianBruniaux/claude-code-plugins · 38 tokens