Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Devlish Review

    1

    Automated Git diff auditor that identifies logic bugs, security vulnerabilities, and broken callers.

    Free

    Security scanned
    devlish-review

    devlish-review

    Example session with this skill installed

    Review protect_env.py and install.sh from this kit for real bugs before I release it.

    • Read your context and instructions
    • Compiled the devlish-review

    I read both files plus settings.json, tested the .env matcher on sample commands, and skipped style opinions.

    Review report

    FileLineCategoryProblemFailing case
    settings.json12SecurityThe hook matcher is Read|Edit|Write|Bash, so Grep and Glob are never checked.A Grep call with path: ".env" returns secret lines and the hook never runs.
    protect_env.py7LogicThe matcher is case-sensitive.cat .ENV is allowed (tested). On a case-insensitive filesystem such as default macOS, that opens .env.
    protect_env.py8SecurityThe regex reads the command text only.cat .e""nv is allowed (tested), and so is python3 -c "open('.'+'env')".
    protect_env.py8LogicOnly one suffix segment is matched after .env.cat .env.production.example is blocked (tested), although it is a template file.
    protect_env.py16Edge caseMalformed hook input raises JSONDecodeError and exits with code 1, which does not block the call.A bad payload lets the tool call through.
    install.sh12LogicAn existing .claude/settings.json is skipped, and only "Skipped" is printed.In a project that already has that file, the .env hook is never installed and the user is not told.

    Next steps

    1. Add Grep|Glob to the matcher and check the path and pattern fields.
    2. Use re.IGNORECASE and block on any hook error (exit 2 in a try/except).
    3. Make install.sh print a clear warning when settings.json is skipped.
    4. Keep saying in the README that this is a safety net and not a guarantee.

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

    About this skill

    Catching real bugs before a commit is hard when you are moving fast. This skill gives your agent a fixed checklist for reviewing your git changes, and tells it to skip style comments.

    What it does

    • Logic: looks for wrong logic and unhandled edge cases in the changed code.
    • Security: looks for unchecked input, exposed secrets and missing auth checks.
    • Callers: checks that anything calling the changed code still works.
    • Tests: points out new behaviour that has no test.

    How it works

    1. The agent runs git diff and git diff --staged to see the changes.
    2. It reads the code around the changes, not only the changed lines.
    3. It checks logic, security, callers and tests, in that order.
    4. It reports each problem with its file, its line and a concrete case that fails.
    5. If nothing is wrong, it says so.

    Why use it

    The skill gives your agent a fixed checklist (logic, security, broken callers, missing tests) and tells it to skip style comments.

    Use cases

    • Check your staged changes before a commit.
    • Check a change before a pull request.
    • Find callers broken by a refactor.
    • Find new code that has no test.

    Known limitations

    • Needs a Git repository.
    • Reads unstaged and staged changes only. It does not compare branches.
    • Gives no feedback on style, formatting or architecture.
    • It is an AI review. It can miss bugs or flag false ones, so it does not replace human review or tests.

    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.

    Requires local Git access. Will not provide feedback on code style or formatting. Requires surrounding code context to be available for full analysis.

    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

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 8 days ago

    • Free with an Agensi account

    Software
    Git

    Permissions

    Terminal / Shell
    Read Files
    Listed8 days ago

    What's inside

    Frequently Asked Questions

    This skill ignores subjective style choices like indentation or naming conventions to focus exclusively on functional risks, such as security vulnerabilities, logic flaws, and breaking changes in your codebase.

    Any language. It works on the git diff and the code around it.

    The skill requires a standard Git environment and access to the local repository; it uses command-line tools to analyze diffs and the surrounding file context without needing external integrations.

    Yes. It reads your unstaged changes (git diff) and your staged changes (git diff --staged). It does not compare branches.

    It is free. You get the review skill, a single SKILL.md that your agent follows. Nothing else is included.

    No, this skill operates as a local auditor that inspects your Git state; it does not require a GitHub account or access to your remote private repositories to function.

    New in Code Quality & Review

    Free