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 houshuang/limbic --skill coordinator-hygienegit clone --depth 1 https://github.com/houshuang/limbicWrote 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/houshuang/limbic/coordinator-hygiene)<a href="https://agentmods.dev/skills/houshuang/limbic/coordinator-hygiene"><img src="https://agentmods.dev/badge/skills/houshuang/limbic/coordinator-hygiene/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/houshuang/limbic/coordinator-hygiene"><img src="https://agentmods.dev/badge/skills/houshuang/limbic/coordinator-hygiene.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 | $0.00115 | $0.01259 |
| Opus 5.5 | $0.00046 | $0.00504 |
| Sonnet 5.5 | $0.00023 | $0.00252 |
| Haiku 4.5 | $0.00012 | $0.00126 |
Grade C, and why
coordinator-hygiene scanned grade C 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 9d 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
description: "Keep a long-running coordinating thread from becoming the bill, and keep its destructive commands from becoming the incident. Use when running a multi-step or multi-agent project from one thread — deciding How it starts
The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Coordinator hygiene
The coordinating thread is usually the most expensive thing in the project. Measured: one thread made 3,838 requests over three days at a median 148K context with a median 204-token output, through 38 compactions — 65% of a project's entire spend, with tool output returned under 2% of the input bill. Another burned 221M tokens on 11 user messages, four of them "continue".
Cost = number of tool calls × standing context. Everything below attacks one of those two factors.
One thread per task; state lives in a file
Write the handoff file first, update it as you go, and start a new thread when
the task changes. HANDOFF.md holds: what is done, what is in flight, the
pinned revision, the next action. Hand off rather than continuing past ~80 tool
calls or ~100k tokens.
"Continue ambitiously" on a 150K-token thread is the single most expensive instruction available, because every subsequent tool call replays that context.
Never poll
Do not wait/sleep/re-check a subagent from a large context — polling cost
32.7M tokens in one thread. Either run the long command in the foreground
with a timeout, or check the artefact the job writes (a file, a ledger row, a
status line), once, when you have reason to think it is done. A background job
that reports on completion is free; a loop that asks is not.
Read narrowly
rgwith line ranges, nevercatof a 672 KB monolith. If lookups are JSON dumps, build the CLI: in one projectMATCHappeared 3 times against 1,174 calls that dumped or grepped the graph.git show/git diffalways with--stator a path. Committed generated site data turns a 9-line fix into a 259 KB diff that then re-enters context.- Never load an image at original resolution into a coordinator. One call was 12 MB. Downscale, or send it to a worker.
- Process docs are not free either: in one repo 33% of patches were HANDOFF, LOG, WORKPLAN and INDEX churn — twice the patch text that reached the data.
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.
- 9d ago First seen · 99 lines · 115 tokens per session scan C 315de88f1d13
coordinator-hygiene is a skill published in the GitHub repository houshuang/limbic (3 stars, last pushed 3d ago), licensed MIT. It adds 115 tokens to every session and 1,259 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-22.
Other skills, from other repositories
sentence-transformers
Framework for state-of-the-art sentence, text, and image embeddings. Provides 5000+ pre-trained models for semantic similarity, clustering, and retrieval. Supports multilingual, domain-specific, and multimodal models. Use for generating embeddings for RAG, semantic search, or similarity tasks. Best for production…
backfill
Reconstruct an OKF bundle by event-sourcing a repository's history (git log and Claude session transcripts). Use when creating an .okf/ bundle for an existing repository that predates this skill, or when resuming an interrupted backfill session. Triggers on: "reconstruct the OKF bundle", "backfill the knowledge…
mstar-branch-worktree
A Morning Star guide for Git branches and worktrees. A worktree is a separate working directory connected to the same Git repository, allowing isolated changes.
mstar-conventions
A guide to Morning Star project folders, plan files, documentation boundaries, and Git branches. It explains where plans, specifications, iterations, knowledge, and project files belong.
memex-sync
Sync Zettelkasten cards across devices via git.
cocosearch-commit
Use when committing changes — analyzes staged (or unstaged) diffs with CocoSearch semantic search and dependency impact to generate comprehensive Conventional Commit messages. Reviews changes thoroughly, then asks user to approve, edit, or abort the commit.