Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    PR Review Gatekeeper

    1

    Turn your agent into a senior code reviewer that blocks risky merges, ranks findings by real impact, and gives authors fixes they can apply in one pass.

    Free

    1 useSecurity scanned
    PR Review Gatekeeper

    PR Review Gatekeeper

    Example session with this skill installed

    Review PR #482 in our billing-api repo before it merges. The author claims the new Redis caching layer around invoice validation is safe and drops p95 latency, but CI only ran the happy path. Check the diff against src/billing/validators, the existing cache helpers in src/lib/cache, and docs/caching.md. I care most about whether validation can be skipped on cache hits, whether the new displayMode config option breaks persistence compatibility, and whether the tests would catch a regression. Give me findings sorted by severity with exact change requests, not vague opinions.

    • Read your context and instructions
    • Compiled the pr review gatekeeper

    Summary

    Reviewed PR #482 (cached invoice validation, 11 files, +540/-72). The bug is real and the intent fits the roadmap, but the warm-cache path still skips validation and the new retry logic ships without failure tests. Verdict: request changes.

    AreaBeforeAfter
    Blocking issuesnone claimedcache-hit writes bypass validateRecord()
    Test coveragehappy path only3 new cases: timeout, success-after-retry, exhausted retries
    API claritydisplayMode also changes persistencerename it or split display/storage flags
    Docsdocs/caching.md untouchedinvalidation note added

    Next steps

    • Route warm-cache writes through the same validation path and add a regression test for the warm-cache write.
    • Add the three retry tests above; assert on outcomes, not implementation internals.
    • Rename displayMode so the name reflects both effects, or split into two settings.
    • Include a migration note for the return-type change, or update all known call sites in this PR.
    • Everything else (naming, guard-clause ordering) batched as optional; not blocking.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Stop merging risky code and low-quality patches. This skill transforms your agent into a disciplined senior code reviewer that prioritizes repository health over contributor convenience. It moves beyond superficial linting to identify architectural mismatches, security vulnerabilities, and hidden maintenance costs.

    What it does

    • Risk Assessment Categorizes findings into Blocking, Major, Minor, and Optional to focus on what matters.
    • Intent Validation Cross-references the diff against PR descriptions, linked issues, and commit messages to ensure the solution matches the problem.
    • API Quality Control Evaluates public interfaces for naming clarity, predictable defaults, and backward compatibility.
    • Maintenance Analysis Identifies fragile logic, premature abstractions, and hidden coupling that increase long-term technical debt.
    • Verification Strategy Demands meaningful tests for failure paths and edge cases, not just happy-path coverage.

    How it works

    1. Context Gathering The agent inspects the PR description, related tickets, and surrounding files before reading the diff.
    2. Impact Analysis It evaluates correctness, design fit, performance, and operational costs.
    3. Structured Feedback The agent generates specific, actionable comments that explain what was observed, why it matters, and how to fix it.
    4. Decision Logic It applies a rigorous checklist to provide a clear verdict: Approve, Request Changes, or Reject with guidance.

    Frameworks & tools

    Works with any programming language or framework. Integrates with Git-based workflows, gh CLI, and standard testing suites like pytest, jest, or go test.

    Why this beats prompting it yourself

    Most prompts result in "LGTM" or pedantic style nits. This skill enforces a skeptical, senior-level posture that catches architectural regressions and security flaws that generic prompts miss.

    Use cases

    • Preventing breaking changes in shared library APIs.
    • Auditing security-sensitive paths in web backend services.
    • Ensuring consistent design patterns across a growing microservices mono-repo.
    • Mentoring junior developers with specific, educational feedback on complex logic.

    Known limitations

    Cannot execute code in sandboxed environments without external tool access. Performance analysis is based on static inspection, not dynamic profiling.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.

    Review depth tracks whether the agent can actually run the project's tests locally; without a working checkout it falls back to static inspection and says so. Teams with unusual conventions may want to add their own style guide alongside it.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    1 use

    Times used to date

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Permissions

    Terminal / Shell
    Read Files

    File Scopes

    src/**
    tests/**
    docs/**
    *.md

    Read access to the checked-out repo plus shell access to run tests is all that is needed; the reviewer never pushes code or opens network connections.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    Unlike generic AI prompts that focus on syntax, this skill adopts a 'skeptical senior' posture to identify architectural flaws, security risks, and technical debt that automated linters and basic LLM queries typically miss.

    It is language-agnostic and works with any Git-based workflow, including GitHub (via gh CLI) and GitLab, provided your AI agent has access to your repository's file system or API.

    You receive the optimized system instructions and decision logic that define the Gatekeeper's behavior, along with a structured checklist for its impact analysis and decision-making process.

    The skill uses static analysis to evaluate code; it does not execute your code or run tests in a sandbox unless your specific agent platform is already configured with external tool access to do so.

    Installation involves importing the skill’s logic into your AI agent’s system prompt or configuration, and ensuring it has the necessary context permissions to read PR descriptions and surrounding file structures.

    The Gatekeeper evaluates the 'intent' by cross-referencing the code diff against your PR descriptions, linked issues, and commit history to ensure the technical solution actually solves the reported problem.

    New in Code Quality & Review

    Free