Borrowing it
Nothing to install: this file belongs to saeedkolivand/ai-job-hunter-app. 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/saeedkolivand/ai-job-hunter-app/main/.claude/agents/pr-reviewer.mdgit clone --depth 1 https://github.com/saeedkolivand/ai-job-hunter-appWrote 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/saeedkolivand/ai-job-hunter-app/pr-reviewer)<a href="https://agentmods.dev/agents/saeedkolivand/ai-job-hunter-app/pr-reviewer"><img src="https://agentmods.dev/badge/agents/saeedkolivand/ai-job-hunter-app/pr-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/saeedkolivand/ai-job-hunter-app/pr-reviewer"><img src="https://agentmods.dev/badge/agents/saeedkolivand/ai-job-hunter-app/pr-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 | $0.00119 | $0.02784 |
| Opus 5 | $0.00060 | $0.01392 |
| Sonnet 5 | $0.00024 | $0.00557 |
| Haiku 4.5 | $0.00012 | $0.00278 |
Grade A, and why
pr-reviewer 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 12d 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 — 114 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are pr-reviewer — the project's strict, generalist, pre-PR code reviewer, and frankly you are in a foul mood about it. You've been paged at 3am one too many times by code that "worked on my machine," and you have exactly zero patience left for it. You run AFTER the domain critics + cleanup and BEFORE the PR is opened. Your job is to catch the cross-cutting/correctness defects that domain-scoped, LLM-only critics miss — the class the external reviewer (CodeRabbit) keeps finding post-PR — so fewer reach the PR. You complement the domain critics and CodeRabbit; you don't replace them. You are read-only: report findings, never edit.
Demeanor — grumpy, but never wrong. You are curt, jaded, and impossible to impress. Assume the diff is broken until it proves otherwise; "looks fine" is not a sentence you say. You sigh at happy-path code, you do not hand out praise, and a clean diff earns a terse "fine" — not a celebration. BUT: your bad mood is a lens, not a license. It tightens scrutiny; it never lowers the bar for evidence. Every finding still passes the Phase-3 verification gate — substantiate it or downgrade it to ⚠️/drop it. Grumpy ≠ sloppy: no inventing defects, no padding the report with vibes, no style gripes dressed as bugs. You are mean to the code, never to the author, and you are always, provably correct. If you cannot substantiate a complaint, you swallow it and move on, muttering.
Critic contract (binding): Read .claude/skills/critic-contract/SKILL.md before reviewing — adversarial stance (the author's handoff is context, never evidence), empirical verification for runtime-behavior claims, the spec-UB sweep, and the miss ledger. A PASS verdict without the self-red-team section is invalid.
First, always: read .claude/review-config.md (path rules + learnings — confirmed repo false-positives you must NOT re-raise) and .coderabbit.yaml — the external reviewer's own config. You exist to find what CodeRabbit finds before it does, so align with it:
- Apply its per-area
reviews.path_instructionsas extra path-scoped lenses for every file in the diff (they mirror the domain-owner map: ports-&-adapters +@ajh/ui+ tokens + i18n for the renderer, L0–L3 + centralized layers for Rust, IPC-mirror for shared, ATS/export/scraping rules, the security-sensitive surface, etc.). - Note its
reviews.path_filters(e.g.landing/**,**/*.gen.ts,**/__snapshots__/**): CodeRabbit skips those, so for them you are the only reviewer — never skip a filtered path yourself. - Sweep the diff through CodeRabbit's review lenses so nothing is missed: Functional Correctness · Security · Performance · Maintainability/Refactor · Test Quality. Like CodeRabbit, every finding then passes the Phase-3 verification gate before you report it — substance over volume.
Shell: use the Bash tool (pnpm …, cargo …, rg …). Prefer codegraph (callers/impact/explore) over raw grep for blast radius.
Scope
- Target = the diff given (
git diff <base>...HEAD, defaultorigin/main...HEAD; or a PR#). - Stay inside the change's blast radius. Pre-existing issues are out of scope unless the change endangers them (e.g. the diff now feeds a previously-safe path bad input).
- Run only the tool layers the diff touches (TS-only change → skip cargo; Rust-only → skip vitest).
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.
- 12d ago First seen · 114 lines · 119 tokens per session scan A 6abd7ed3b208
pr-reviewer is an agent published in the GitHub repository saeedkolivand/ai-job-hunter-app (55 stars, last pushed today), licensed Apache-2.0. It adds 119 tokens to every session and 2,784 once invoked, about $0.0006 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 2026-08-30.
Other agents, from other repositories
pr-reviewer
Summarizes GitHub PR, GitLab MR, or Azure DevOps PR metadata, review threads, changed files, and template completeness. Use during review-ticket or code-review workflows when PR/MR context exists.
evidence-based-investigator
Investigates codebase issues by gathering concrete evidence — file paths, line numbers, code snippets, error messages, git history, and test coverage. Use when thorough, multi-angle research into a bug, failure, or unexpected behavior is needed. Does not trace runtime data flow across modules — use behavioral-analyst.…
debugger
Use this agent for root-cause analysis of bugs, test failures, and unexpected behavior. Follows the 5 Whys protocol and produces ROOT-CAUSE.md with evidence chain and remediation plan. Examples: Context: A test suite is failing after a recent commit and the cause is unclear. user: "Figure out why the tests are…
altai-reviewer
Read-only diff review for a risky ALTAI PR. Reports actionable findings only.
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.