Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/Jerry0022/dotclaudenpx agentmods add skills/jerry0022/dotclaude/mcp-reapWrote 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/jerry0022/dotclaude/mcp-reap)<a href="https://agentmods.dev/skills/jerry0022/dotclaude/mcp-reap"><img src="https://agentmods.dev/badge/skills/jerry0022/dotclaude/mcp-reap/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/jerry0022/dotclaude/mcp-reap"><img src="https://agentmods.dev/badge/skills/jerry0022/dotclaude/mcp-reap.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00029 | $0.00920 |
| Opus 5.5 | $0.00012 | $0.00368 |
| Sonnet 5.5 | $0.00006 | $0.00184 |
| Haiku 4.5 | $0.00003 | $0.00092 |
Grade A, and why
mcp-reap 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 1mo 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.
How it starts
The opening of the file, as written. The whole thing — 74 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MCP Reap
Reclaims orphaned Claude Desktop MCP server child processes — leftovers from previously-closed sessions that were never reliably terminated and linger, accumulating RAM. Driven by two hooks so it runs both once per session and periodically thereafter, without requiring a restart:
hooks/session-start/ss.mcp.reap.js— fires once, everySessionStart.hooks/stop/stop.mcp.reap.js— fires onStop(end of a response turn), gated by a ~20-minute per-worktree cooldown so it doesn't scan every turn. The cooldown marker uses the same atomic write-temp-then- rename pattern ashooks/lib/session-id.js, keyed to an md5 hash ofprocess.cwd()inos.tmpdir().
Both hooks spawn scripts/mcp-reap.js --apply --json detached,
fire-and-forget, windowless ({ detached: true, stdio: 'ignore', windowsHide: true } + .unref()) and exit immediately — reaping never
delays a session start or a response turn.
Safety model
The scan-and-kill logic lives in hooks/lib/mcp-reaper.js (fully unit
tested, 46/46 in mcp-reaper.test.js) and is Windows-only — the
dead-parent-PID orphan signal only means anything on Windows, so the module
is a documented no-op on macOS/Linux. A process is only ever a reap
candidate when it matches an MCP-server launcher signature, its parent PID
is confirmed dead, AND it falls outside the live-Claude census (every
currently-live claude/claude.exe process plus its full descendant
subtree) and the caller's own process subtree — an empty/unbuildable census
means "protection unknown" and refuses to produce any candidates at all.
Every candidate is re-validated against a fresh process snapshot immediately
before both the SIGTERM and any SIGKILL escalation (TOCTOU-safe), so a pid
reused for something else between scan and kill is skipped, never touched.
The reap() module itself defaults to dry-run — nothing is ever terminated
unless a caller explicitly passes --apply; both hooks opt into it
deliberately, on the reasoning that a leaked orphan is safe to reclaim
automatically once the safety net above holds. Any enumeration or kill
failure fails safe (skip, never throw).
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.
- 1mo ago Changed e30c8e07c41a
- 2mo ago First seen · 74 lines · 29 tokens per session scan A 05ae761f8dd4
mcp-reap is a skill published in the GitHub repository Jerry0022/dotclaude (6 stars, last pushed 11d ago), licensed MIT. It adds 29 tokens to every session and 920 once invoked, about $0.0001 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 2026-08-31.
mcp-reap compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| mcp-reapJerry0022/dotclaude | Skill | 6 | 11d ago | MIT | A | 29 | Claude Code |
| imagegenopenai/codex | Skill | 133k | yesterday | Apache-2.0 | A | 113 | Codex |
| playwright-climicrosoft/playwright | Skill | 99k | yesterday | Apache-2.0 | A | 19 | Claude Code |
| local-ai-agentsmicrosoft/ai-agents-for-beginners | Skill | 79k | 4d ago | MIT | A | 200 | Claude Code, Codex |
| systematic-debuggingobra/superpowers | Skill | 313k | 3d ago | MIT | A | 21 | Claude Code |
| next-cache-components-adoptionvercel/next.js | Skill | 143k | yesterday | MIT | A | 95 | Claude Code |
| next-partial-prefetching-adoptionvercel/next.js | Skill | 143k | yesterday | MIT | A | 103 | Claude Code |