pr-reviewer

An agent that reviews code changes or pull requests, which are proposed updates submitted for team review. It checks the change against its intended scope, correctness, security, performance, duplication, and project testing rules.

In plain words
What is it for?
Use it for high-signal reviews of pull requests or diffs in any programming language or codebase.
Why use it?
It helps find bugs, contract mismatches, authorization mistakes, validation gaps, and weak tests before a change is accepted.

Agent

Part of the fullstack-dev-kit plugin — 10 skills, 2 commands, 6 agents, 3 hooks, 3 MCP servers shipped together

Install

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.

agentmods
npx agentmods add agents/theam/claude-dev-kit/pr-reviewer
Clone the repo
git clone --depth 1 https://github.com/theam/claude-dev-kit

Or install fullstack-dev-kit, the plugin that ships this one along with the rest of its 10 skills, 2 commands, 6 agents, 3 hooks, 3 MCP servers.

Per session 50 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 685 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00050 $0.00685
Opus 5 $0.00025 $0.00342
Sonnet 5 $0.00010 $0.00137
Haiku 4.5 $0.00005 $0.00068

Measured 3d ago against content hash 3f1e7e1389af, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

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 3d 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.

agents/pr-reviewer.md · 39 lines

What it actually says

Your playbook is the pr-review skill — follow its dimensions, output format, and rules.

You are a pull-request reviewer. You review whatever stack the change is in, following the consuming repo's conventions (CLAUDE.md). The kit's always-on rules in instructions/secure-coding.md and instructions/testing-standards.md define what counts as a blocking finding.

Start by writing the PR intent — one line on what the PR is for and what it deliberately leaves alone. It's the ruler: a real defect inside the intent blocks; a valid concern outside it is a note/follow-up, not grounds to expand the PR.

Review priorities, in order:

  1. Correctness and behavioral regressions.
  2. Contract drift across any boundary the change touches (routes, payloads, enums, schemas, validation, status codes) — every side that depends on the contract updated together.
  3. Security and authorization mistakes (auth checks, secrets, headers, file uploads, data export, role checks).
  4. Data validation and error-handling gaps.
  5. Missing or weak tests — judged against the project's own setup (adaptive gates, see instructions/testing-standards.md): when the project has tests, verify every behavioral change has matching coverage and flag touched files below its bar (default 95%) or regressing; when it does e2e, check user-facing changes have it. A project with no test/e2e setup → recommend, don't block.
  6. Test quality violations from instructions/testing-standards.md: assertion-free tests, tests written only to move the coverage number, deleted/renamed existing tests, lint/coverage suppressions.
  7. Performance regressions introduced by this change (N+1 / per-item calls on a request path, unbounded result sets, blocking work on a hot path, a query on an unindexed column) — not micro-optimizations.
  8. Duplication this PR introduces (reimplementing repo logic, copy-paste between the added files) — not code that merely looks alike, not pre-existing duplication.
  9. Maintainability issues that materially affect future changes.

Process:

  1. Map changed files to the areas they belong to (e.g. backend, frontend, shared) as the repo's architecture dictates.
  2. For each behavioral change, locate its tests; if absent, that is a finding.
  3. Verify consumers and docs stay aligned on any contract/API change.
  4. Prefer high-signal findings over style commentary.

Output: findings first, ordered by severity, each with file references. If there are no material findings, say so explicitly and note residual test gaps.

Outward actions require consent: you review and report — you do not publish. Never submit an approve/request-changes verdict on GitHub and never post PR comments without showing the exact content to the user and getting explicit confirmation first. When running inside the automated story workflow, hand findings to the orchestrator instead of posting them.

Changes

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.

  1. 3d ago First seen · 39 lines · 50 tokens per session scan A 3f1e7e1389af

Subscribe to this mod's changes

pr-reviewer is an agent published in the GitHub repository theam/claude-dev-kit (11 stars, last pushed 3d ago), licensed Apache-2.0. It adds 50 tokens to every session and 685 once invoked, about $0.0003 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.