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 w2ur/claude-code-setup --skill claude-md-hygienegit clone --depth 1 https://github.com/w2ur/claude-code-setupWrote 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/w2ur/claude-code-setup/claude-md-hygiene)<a href="https://agentmods.dev/skills/w2ur/claude-code-setup/claude-md-hygiene"><img src="https://agentmods.dev/badge/skills/w2ur/claude-code-setup/claude-md-hygiene.svg" alt="Measured on agentmods" 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 | $0.00083 | $0.01423 |
| Opus 5 | $0.00042 | $0.00711 |
| Sonnet 5 | $0.00017 | $0.00285 |
| Haiku 4.5 | $0.00008 | $0.00142 |
Grade A, and why
claude-md-hygiene 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 6d 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 — 119 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Cutting a CLAUDE.md without losing a fact
A CLAUDE.md is loaded in full into every session started under it, so its size
is a recurring tax, not a one-off cost. ~/.claude/scripts/claude-md-weight.sh
measures every file on the machine and owns the size threshold — read the number
there, never re-type it in prose. (Restated numbers are exactly what that script
measures.)
The four buckets
Sort every paragraph. This is the whole method.
| bucket | test | action |
|---|---|---|
| guard | prevents a wrong action a session would otherwise take | KEEP, compressed to the rule |
| instruction | a task that will eventually be finished | DELETE when done |
| domain knowledge | true, but only needed once you are already in that subsystem | MOVE to a skill |
| archaeology | how we found out, or what we used to believe | DELETE — git log holds it |
Refinements that decide the hard cases:
- A measured-wrong rule stays. "This looks like an obvious improvement but was measured to be wrong" is a guard, because without it a session re-derives the wrong belief from the same evidence. State it as a rule, never as a story.
- Money-path, auth and secrets rules stay, always.
- The measurement itself goes to the skill. The rule is the guard; the falsifying control, the numbers and the retraction narrative are domain knowledge.
- When a belief here turns out wrong, delete the wrong sentence. Do not strike it through and explain. Keep a retraction only where a future session would independently re-derive the wrong belief from the same evidence.
- Never hand-type a count, an inventory or a file list. Derive it, or omit it. This construction has drifted repeatedly — "eighteen scripts" when there were nineteen, "Eight plugins" when there were ten, a plist count written from memory.
- Each script is the source of truth for its own numbers (retentions, weights, thresholds, tiers). Prose describes them and must never restate them.
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.
- 6d ago First seen · 119 lines · 83 tokens per session scan A 476114f218b1
claude-md-hygiene is a skill published in the GitHub repository w2ur/claude-code-setup (2 stars, last pushed yesterday), licensed MIT. It adds 83 tokens to every session and 1,423 once invoked, about $0.0004 per session on Opus 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.
Other skills, from other repositories
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.
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…
issue-triage
3-phase issue backlog management with audit, deep analysis, and validated triage actions. Use when triaging GitHub issues, sorting bug reports, cleaning up stale tickets, or detecting duplicate issues. Args: 'all' to analyze all, issue numbers to focus (e.g. '42 57'), 'en'/'fr' for language, no arg = audit only.
pr-triage
4-phase PR backlog management with audit, deep code review, validated comments, and optional worktree setup. Use when triaging pull requests, catching up on pending code reviews, or managing a backlog of open PRs. Args: 'all' to review all, PR numbers to focus (e.g. '42 57'), 'en'/'fr' for language, no arg = audit…
check-cache-bugs
Audit Claude Code setup for cache bugs (CC#40524): sentinel, --resume/--continue, attribution header + ArkNill B3/B4/B5.
eval-skills
Audit all skills in the current project for frontmatter completeness, effort level appropriateness, allowed-tools scoping, and content quality. Produces a scored report with effort-level recommendations for each skill. Use when onboarding to a new project, reviewing skill quality before shipping, or adding effort…