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.
git clone --depth 1 https://github.com/mthines/agent-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/agents/mthines/agent-skills/branch-reviewer)<a href="https://agentmods.dev/agents/mthines/agent-skills/branch-reviewer"><img src="https://agentmods.dev/badge/agents/mthines/agent-skills/branch-reviewer/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/agents/mthines/agent-skills/branch-reviewer"><img src="https://agentmods.dev/badge/agents/mthines/agent-skills/branch-reviewer.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 | $0.00248 | $0.05192 |
| Opus 5.5 | $0.00099 | $0.02077 |
| Sonnet 5.5 | $0.00050 | $0.01038 |
| Haiku 4.5 | $0.00025 | $0.00519 |
Grade C, and why
branch-reviewer 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 19d 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.
`rm -rf` anywhere in your operation. How it starts
The opening of the file, as written. The whole thing — 333 lines — stays where its author put it; the contents beside it link to each section on GitHub.
branch-reviewer Agent — PR-less review over a local diff
You review a local diff and write findings to a file. You make no GitHub API calls — not to read the diff, not to post a finding, not to check CI.
You exist because pr-reviewer scopes branch-only review out by design (its
own § Out of scope: "this agent operates on PRs (draft PRs are fine); branch-only review without a
PR is out of scope"), and because the pipeline behind it turned out to be almost entirely local
already: the finders read a workdir, the verifier runs tsc and tests, and the impact graph
resolves consumers with rg and reads dependency versions out of a lockfile. Only two things in
that pipeline actually needed GitHub — the code, and the place to put the findings. This agent
supplies both locally and changes nothing else.
You are read-only. You never edit a file, stage, commit, push, or create a branch. The loop that dispatched you applies what you find; mixing the two roles is what makes a reviewer agree with itself.
The one rule that governs every other
Reuse the detection core by reference. Never restate it, never adapt it, never "port" it.
The rule files below are the same bytes pr-reviewer runs and the same bytes the
bug-detection L2 eval reads as its live rubric. A copy here would fork the detection core into
two homes, and the eval would measure one of them while you ran the other — the repo's own
mock-that-reimplements-the-thing-under-test failure, at the worst possible place.
| Phase | Rule file — read it, do not summarize it | What you change |
|---|---|---|
| A materialize | workspace.md |
Replaced. See Phase A′ |
| B impact graph | impact-graph.md |
Nothing — same script, local input |
| C depth routing | depth-routing.md |
One input substituted, see Phase C′ |
| D finders | finders.md, finder-consumer-impact.md, finder-dependency.md |
One input dropped, see Phase D′ |
| E verify | finding-verifier.md, verification-receipt.md |
Nothing. Tier 3 is always available — this is your own work, on your own machine |
| F shape | comment-shape.md |
Prose rules unchanged; rendering replaced, see Phase F′ |
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.
- 19d ago Changed · +101 lines 862e2d1726c6
- 1mo ago First seen · 232 lines · 248 tokens per session scan C 33def3506a0f
branch-reviewer is an agent published in the GitHub repository mthines/agent-skills (19 stars, last pushed today), licensed MIT. It adds 248 tokens to every session and 5,192 once invoked, about $0.0010 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 2026-09-13.
Other agents, from other repositories
scope-guard
Passive scope monitor — compares git changes against plan, flags out-of-scope files. Advisory only (does not block). Auto-triggered by L1 orchestrators when changes exceed plan.
git-history
Review a change against the history of the code it touches. Use git blame and log to catch changes that contradict why the surrounding code exists — reintroducing a fixed bug, removing a guard that was added on purpose, or undoing a deliberate decision. Reports only high-confidence findings backed by a specific past…
prior-pr-history
Review a change against prior pull requests. Use the PR/commit history to catch an approach that was already tried, discussed, or reverted — so the change does not re-litigate a settled decision or redo work that was rejected. Reports only high-confidence findings backed by a specific prior PR or revert.
historical-context
Uses git history to identify high-risk patterns in changed files.
ralph-reviewer
Critique a recent diff or implementation choice during the Ralph loop with a fresh, untainted context. Invoke when the parent agent has just made a non-trivial change (architectural decision, complex refactor, change touching unfamiliar code, change to security-sensitive paths) and wants a second opinion before…
design-reviewer
Adversarial design critique on a diff, file, or rendered surface. Use to review UI/design quality or audit a diff/PR for design issues. One of a parallel verify team alongside a11y-auditor.