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 closedloop-ai/claude-plugins --skill repo-hardeninggit clone --depth 1 https://github.com/closedloop-ai/claude-pluginsWrote 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/closedloop-ai/claude-plugins/repo-hardening)<a href="https://agentmods.dev/skills/closedloop-ai/claude-plugins/repo-hardening"><img src="https://agentmods.dev/badge/skills/closedloop-ai/claude-plugins/repo-hardening/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/closedloop-ai/claude-plugins/repo-hardening"><img src="https://agentmods.dev/badge/skills/closedloop-ai/claude-plugins/repo-hardening.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 | $0.00184 | $0.02132 |
| Opus 5.5 | $0.00074 | $0.00853 |
| Sonnet 5.5 | $0.00037 | $0.00426 |
| Haiku 4.5 | $0.00018 | $0.00213 |
Grade A, and why
repo-hardening 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 7d 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 — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Repo hardening
Audit a repository against a small set of engineering principles, then hand back a plan the owners can act on in small reviewable units. The plan is the deliverable. In an audit, do not change code, open pull requests, or file tickets unless the user explicitly asks after reading the plan. Direct use changes code only when the user's request is for that change.
For rigorous work on one task instead of a whole-repo audit, skip to Direct use.
Harness notes
- Works in any agent harness. Paths like
references/...andassets/...are relative to this skill's folder, not the repository under audit; nothing here depends on a home directory layout. - "Subagent" below means a parallel worker or a fresh session if your harness has one. If it cannot spawn workers, do the areas sequentially, keep notes in files in the system temp directory, and run the self-review as a separate pass, recording that it was not independent.
- "Checklist" means your harness's todo tool if it has one, otherwise a scratch file you keep updated.
- Asking the user means one short question in the conversation, whatever tool the harness provides for that.
Non-negotiables
- Every finding carries evidence in the same sentence: a
path:line, a commit, a PR or review comment, or a command and its output. Mark inference as inference. Never invent callers, history, or rules. - Read-only, for the analyzed repo. Subagents that investigate get read-only tools for it; where the harness cannot restrict tools, the brief's FORBIDDEN line is the restriction. Each may write exactly one file, its own scratch notes file outside the analyzed repo; a worker whose tools cannot write returns its notes in its final reply and the lead writes the file. Nothing in the analyzed repo is modified or executed apart from git and read commands (no installs, builds, or test runs), and the plan is written outside the analyzed repo's directory unless the user names a path inside it.
- The repo's own rules win. Where the instruction files of the repo being analyzed or changed (AGENTS.md, harness-specific agent instruction files, contributing guides) conflict with a default in this skill, such as the refactor playbook's no-shims step or the autonomy defaults, follow the repo. In an audit, a finding whose fix a repo rule forbids, for example deleting a compatibility shim the repo says to keep, goes to the decisions for a human, not into a unit. In direct use, follow the rule and name the conflict in the reply.
- Prefer the strongest prevention that fits: architecture or types, then static checks and source gates, then guard tests, then a rule in the owning instruction file, then advisory notes. Rules that nothing enforces are findings, not fixes. See trust-ladder.md.
- Subtract before you add. A plan that only adds rules, tools, and process has missed the deletions.
- Outline, not spec. Each unit of work gets a goal, evidence, the mechanism, where it lives, and how to prove it. Owners fill in the rest.
- Guard the context window. Bulk reading happens in subagents that return pointers and short findings, never file dumps.
What ships with it
8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 7d ago Changed · +1 lines · +184 tokens per session eaf17e4e48a1
- 9d ago First seen · 75 lines · 0 tokens per session scan A e3044cc065b0
repo-hardening is a skill published in the GitHub repository closedloop-ai/claude-plugins (122 stars, last pushed today), licensed Apache-2.0. It adds 184 tokens to every session and 2,132 once invoked, about $0.0007 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 2026-10-11.
Other skills, from other repositories
qa-refactoring
Refactors code while preserving behavior. Use when reducing technical debt, planning codemods, applying strangler migrations, or tightening CI guardrails around risky changes.
check-work
Check your work with a verification subagent that reviews diffs, runs builds and tests, and evaluates correctness. Read this file for instructions. Use when asked to "check work", "verify changes", "self-verify", "/check-work", "/check", "/verify", or "/self-verify".
review
Review code changes, pull requests, patches, or a scoped code area for actionable correctness, security, compatibility, and test risks with file and line evidence. Use for review or audit requests; do not use for general proofreading, feature implementation, or debugging a reported failure when the user wants a fix.
debug
Diagnose a concrete failure, regression, crash, hang, flaky test, or incorrect runtime behavior by reproducing it, testing hypotheses, and identifying the evidence-backed root cause. Use when symptoms or failing output exist; do not use for feature implementation without a failure, general code review, or a conceptual…
adversarial-review
Run a file-audited completion loop that requires two consecutive fresh-context subagents to independently try to disprove the same versioned, digest-checked task result. Use when the user invokes $adversarial-review, asks for two independent adversarial approvals, or requires a result to survive repeated attempts to…
g-review
Run the review gate on the current branch diff. Runs the test suite, captures the diff, and dispatches code-lead, which verifies done conditions and reviews the diff itself. Issues MERGE READY or HOLD.