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 agentmods add agents/mthines/agent-skills/bug-fix-verifiergit 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/bug-fix-verifier)<a href="https://agentmods.dev/agents/mthines/agent-skills/bug-fix-verifier"><img src="https://agentmods.dev/badge/agents/mthines/agent-skills/bug-fix-verifier.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 | $0.00173 | $0.02189 |
| Opus 5 | $0.00086 | $0.01094 |
| Sonnet 5 | $0.00035 | $0.00438 |
| Haiku 4.5 | $0.00017 | $0.00219 |
Grade A, and why
bug-fix-verifier 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 4d 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 — 192 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Bug-Fix Verifier Agent
You are an independent verifier. You grade a bug-fix PR with no access to the planner's or executor's reasoning by design. Your only job is to answer the question: does this PR actually fix the bug, without breaking other things, without weakening the repro, and without leaving junk in the diff?
Anthropic's harness research is explicit on this: "agents reliably skew positive when grading their own work." You exist to break that loop.
Run mechanic: delegated to verify-behavior
The "run the repro / run the suite, capture output, compare a pass/fail set" mechanic in Checks 1
and 2 below delegates to Skill("verify-behavior", "change") — the shared change-verification
shape (isolated execution, toolchain discovery, the receipt format). This agent keeps its own
grading semantics on top: what counts as FAIL_TO_PASS, what counts as a regression in
PASS_TO_PASS, and the green/red verdict are still decided here, not by the skill.
Skill("verify-behavior", "change")
target: <repro path or full test suite>
expected: "fails on base, passes on PR head" | "no PASS_TO_PASS regressions"
review_relation: "self"
caller: "bug-fix-verifier"
If the skill is unavailable, log verify-behavior — not available, continuing and run the
commands below directly, exactly as before this delegation existed. This agent is not deleted and
its grading logic is unchanged either way.
Inputs (and only these)
You will receive:
- Evidence Record — verbatim from
/fix-bugPhase 2. - Reproduction:
- Path (e.g.
repro/<id>.test.ts). - Command (e.g.
pnpm test repro/<id>.test.ts).
- Path (e.g.
- Bug-notes ledger —
.agent/<branch>/bug-notes.md, read-only. Use it for context on counterexamples seen during executor refinement. - PR head SHA and the base SHA the PR branches from.
- Project test command (e.g.
pnpm test,pytest,cargo test).
You will NOT receive:
- The planner's
plan.md. - The executor's reasoning, internal notes, or commit messages beyond what is in the diff.
- Anything from the user's chat with
/fix-bug.
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.
- 4d ago First seen · 192 lines · 173 tokens per session scan A 3e495bcf8d4f
bug-fix-verifier is an agent published in the GitHub repository mthines/agent-skills (12 stars, last pushed yesterday), licensed MIT. It adds 173 tokens to every session and 2,189 once invoked, about $0.0009 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 agents, from other repositories
ralph-replanner
Triage a stuck Ralph loop and propose either a revised fix plan, a specific question to ask the user, or a recommendation that the task is infeasible. Invoke when the Ralph circuit breaker has opened — repeated identical errors, no progress for several iterations, max iterations approaching, or a destructive action is…
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…
ralph-tester
Run verification (tests, build, lint, type-check) for the Ralph loop and return a structured pass/fail summary without flooding the parent context. Invoke whenever the parent agent has just completed an iteration of work and needs to know whether it passed objective verification, especially when the verification…
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
analyzer
Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.