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.
npx skills add RickyTong1/audit-harness --skill audit-recovergit clone --depth 1 https://github.com/RickyTong1/audit-harnessWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/rickytong1/audit-harness/audit-recover)<a href="https://agentmods.dev/skills/rickytong1/audit-harness/audit-recover"><img src="https://agentmods.dev/badge/skills/rickytong1/audit-harness/audit-recover/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/rickytong1/audit-harness/audit-recover"><img src="https://agentmods.dev/badge/skills/rickytong1/audit-harness/audit-recover.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00105 | $0.00952 |
| Opus 5.5 | $0.00042 | $0.00381 |
| Sonnet 5.5 | $0.00021 | $0.00190 |
| Haiku 4.5 | $0.00011 | $0.00095 |
Grade A, and why
audit-recover scanned grade A with 0 findings 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 2mo 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
/recover — Context 恢复
触发方式
/recover # 恢复当前 session 的 context
/recover full # 恢复当前 session + 最近 3 天的完整 context
/recover {session_id} # 恢复指定 session 的 context
核心原则
不确定就查审计记录,不要凭印象工作。
这不是一个"可选的辅助工具"——当 context 丢失时,/recover 是唯一可靠的恢复手段。 凭印象工作会导致重犯已被用户否定的错误,这比不工作更糟糕。
何时触发
用户手动触发
用户说"你刚才说的什么来着"、"回忆一下之前的讨论"等。
Claude 主动触发(更重要)
| 信号 | 判断方式 |
|---|---|
| 无法回忆会话前期的具体内容 | 用户引用了"之前讨论的",但你找不到对应内容 |
| 不确定当前任务做到哪一步了 | 要继续工作但不知道上次停在哪 |
| 要提出的方案可能已被用户否定 | 不确定某个判断是否被纠正过 |
| 回答变得笼统,缺少具体细节 | 无法说出具体的数字、文件名、结论 |
执行步骤
分级加载
按以下优先级渐进加载,避免一次性占满 context:
Level 0 — 索引摘要(~200 tokens)
读取 .claude/runs/index.json → 最近 5 条 entry 的 id + task + status
Level 1 — 任务状态(~500 tokens)
读取 .claude/runs/.current_session → 获取当前 session_id
读取 .claude/runs/{session_id}/session.json → task + start_time
读取 .claude/runs/{session_id}/audit_trail.jsonl → 最后 3 条记录的 action + output
→ "当前做到哪了"
Level 2 — 用户修正 + 决策上下文(~1,000 tokens)⚠️ 最重要
在 .claude/runs/ 下所有 audit_trail.jsonl 中搜索 "user_correction"
搜索所有 conclusions 字段
读取最新 .claude/runs/daily/ 下的日报 §四异常 + §六待办
→ "用户否定了什么、得出了什么结论、有什么未解决的问题"
Level 3 — 完整上下文(~3,000 tokens,仅在 /recover full 时加载)
完整的 audit_trail.jsonl
完整的 audit_pending.jsonl(如果还有未归档的)
输出恢复报告
=== Context 恢复报告 ===
来源: .claude/runs/{session_id}/
恢复级别: Level 2
当前任务: {session.task}
任务状态: {最后一个 audit 记录的 output}
⚠️ 用户修正记录(必须遵守):
1. [时间] 原始判断: "..."
修正: "..."
原则: ...
关键结论:
- ...
未解决问题:
- ...
⚠️ 以上内容从审计记录恢复,不是从记忆中回忆的。
注意事项
- 恢复的 user_correction 记录具有最高优先级——后续工作中不得违反
- 如果恢复后发现当前想法与 user_correction 矛盾,必须修正想法
- 恢复报告应简洁,避免把整个审计记录复制到 context
- 优先恢复"为什么"和"不要做什么",而非"做了什么"
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.
- 2mo ago First seen · 109 lines · 105 tokens per session scan A cc1e493fe3e3
audit-recover is a skill published in the GitHub repository RickyTong1/audit-harness (468 stars, last pushed 5mo ago), licensed MIT. It adds 105 tokens to every session and 952 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
audit-recover compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| audit-recoverRickyTong1/audit-harness | Skill | 468 | 5mo ago | MIT | A | 105 | Claude Code |
| knowledge-ingestopenclaw/openclaw | Skill | 394k | today | MIT | A | 80 | Claude Code, Codex |
| mem0-oss-to-platformmem0ai/mem0 | Skill | 69k | yesterday | Apache-2.0 | A | 217 | Claude Code, Codex |
| forgetasgeirtj/system_prompts_leaks | Skill | 71k | today | CC0-1.0 | A | 0 | Claude Code, Codex |
| Cortexdanielmiessler/LifeOS | Skill | 20k | 2d ago | MIT | A | 184 | Claude Code |
| memoryopensquilla/opensquilla | Skill | 7.1k | yesterday | Apache-2.0 | A | 44 | Claude Code, Codex |
| ha-data-storesshiwenwen/hope-agent | Skill | 1.9k | today | MIT | A | 115 | Claude Code, Codex |