Borrowing it
Nothing to install: this file belongs to ssheleg/sshlg-skills. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/ssheleg/sshlg-skills/main/CLAUDE.mdgit clone --depth 1 https://github.com/ssheleg/sshlg-skillsWrote 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/instructions/ssheleg/sshlg-skills/claude-md)<a href="https://agentmods.dev/instructions/ssheleg/sshlg-skills/claude-md"><img src="https://agentmods.dev/badge/instructions/ssheleg/sshlg-skills/claude-md/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/instructions/ssheleg/sshlg-skills/claude-md"><img src="https://agentmods.dev/badge/instructions/ssheleg/sshlg-skills/claude-md.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 | $0.01807 | $0.01807 |
| Opus 5 | $0.00903 | $0.00903 |
| Sonnet 5 | $0.00361 | $0.00361 |
| Haiku 4.5 | $0.00181 | $0.00181 |
Grade A, and why
sshlg-skills CLAUDE.md 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 10d 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 — 121 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — sshlg-skills
House rules for this repository only. Global doctrine (language, quality
bar, ops autonomy, skill routing) already loads from ~/.claude/CLAUDE.md in
the same session; repeating it here would create a second source of truth that
drifts. What follows is what is true here and nowhere else.
What this repo is
The umbrella of the ssheleg skill family: a zero-dependency Node launcher plus eight skills as pinned git submodules. It ships no doctrine of its own — each member carries its own — but it owns the family's routing block, the one piece of the family that writes into a file the operator owns and did not write.
That last sentence is the whole risk profile. ~/.claude/CLAUDE.md has no
version control behind it. Everything below exists because of it.
The gate
npm test # test/validate.py, then every discovered test/*_test.js
Ratchets live in docs/DOCMAP.md and are counted by running that command, never
carried across from a previous edit. python3 test/check_pins.py is deliberately
outside it: it queries the npm registry, and npm test must work offline.
Invariants — each one has a check, and the check has been watched failing
- One channel per agent. A plain copy under
~/.claude/skills/<id>shadows the plugin of the same name and serves its frozen version forever.installandupdateprune those copies after every skills-CLI run. - An address a live document claims resolves, or the commit does not land. This
repository's README is an argument about evidence carrying its own receipt, and it told a
reader to run a
test:negativesscript —task-pipeline's, andMissing scripthere — beside a convergence file naming ascripts/check that has never existed in this tree (UM-03, requirement M-07). Name a dead command, never claim it: a document that quotes one as runnable is refused by this guard, correctly, and both repairs above are written that way. Neither was found by reading them; both were found by resolving all 83 addresses in the five live documents at once.test/doc_refs.pyextracts and resolves,test/validate.pyowns the corpus, and four plants in CI cover the three claim classes plus the empty corpus that would pass everything. The dated records are counted, not gated, and the count prints on every run — their rows cite member repositories and states that were true at a commit, and rewriting those is a thing this repository has already decided against. - Submodule urls are HTTPS.
gh repo create --source .sets an SSH remote andgit submodule addinherits it; everything stays green on the machine that has the key and the release smoke test exits 128 everywhere else.test/validate.pyrefuses a non-HTTPS url, with a negative self-test in CI. - The pin is the promise. A checkout of any hub commit must install exactly
the versions
skills.jsonadvertises, so the validator reads the version out of the submodule rather than trusting.gitmodules. Three places move together:skills.json, the submodule pointer, the README table. - The operator's wording wins — and now says when it has diverged.
authoredentries take precedence over the packaged text on every run. Since that means a reworded router never arrives,routersreports drift and--adoptis the only path that replaces a person's words, one router at a time, parking what it replaced. - A guard reads what would RUN, not what a payload contains. The hygiene guard
refused any Bash text carrying
skills update <member>— including a document quoting it, which blocked a verification-ledger row from being committed.executablePart()now drops heredoc bodies fed to a non-shell and whole-line comments, and keeps everything else: abash <<EOFbody still runs, and quoted strings are still read, becausebash -c '…'is a real invocation. Closing the false positive turned up the matching bypass —bareNamekept the trailing quote, so a genuine quoted invocation had been passing untouched. - A guard decides in a pure module; the hook only moves bytes. Every refusal
this pack can make is a function of a payload (
lib/guard.js,lib/hygiene.js,lib/repogate.js), fixtured without aHOME, and the filesystem appears only where a backup is actually taken. Nothing depends on a hook entry'siffilter: the reference states it is best-effort and fails open on a command it cannot parse, so a guard resting on it ships with a documented bypass. - A hook fails silent, and a refusal names its remedy. A guard that throws
breaks every turn in every session, including sessions of packs that never asked
for this one; a refusal with no next step is how an operator learns to switch a
hook off. Both are asserted in
test/hooks_e2e_test.js, which runs the real scripts as processes.
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.
- 10d ago First seen · 121 lines · 1,807 tokens per session scan A f78f8f6b8bb1
sshlg-skills CLAUDE.md is an instructions file published in the GitHub repository ssheleg/sshlg-skills (6 stars, last pushed today), licensed MIT. It adds 1,807 tokens to every session, about $0.0090 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 instructions, from other repositories
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.