Parrot Review

    1

    Focused code review with six engineering profiles, actionable findings, and explicit verdicts.

    Free

    2 usesSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Parrot Review

    Example session with this skill installed

    Synthetic demonstration fixture, not a production repository. Review this Python change under the google profile. Intended behavior: an empty list returns 0; otherwise return the arithmetic mean. No tests are provided.

    --- a/stats.py
    +++ b/stats.py
    @@ -1,2 +1,2 @@
     def average(values):
    -    return sum(values) / len(values)
    +    return sum(values) / (len(values) - 1)
    
    • Read your context and instructions
    • Compiled the parrot review

    Checked demonstration

    Synthetic input, reviewed during package preparation. The output below was corrected after review; it is not a benchmark or a claim about a production repository.

    Scope: Supplied synthetic stats.py diff only
    Profile: Google
    Context limitations: No surrounding code or tests were supplied. No independent reviewer context was available within the demo run.
    Checks executed: None, static review only

    FINDINGS

    • stats.py:2 - The divisor len(values) - 1 does not calculate the arithmetic mean for nonempty lists. A one-element list divides by zero; lists with more than one element use the wrong divisor.
    • advisory: No tests were provided. Add tests for the empty-list result and representative nonempty lists.

    REVIEW VERDICT: FAIL

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

    About this skill

    Parrot Review

    A free, portable adaptation of the one-shot code-review skill from my open-source Parrot Claude Code plugin.

    Review a diff for correctness, error handling, interface compatibility, and long-term code health without a stream of formatting nits. Findings include evidence and a PASS, FAIL, or INCOMPLETE review verdict.

    Included

    • Six review profiles: Google, Meta, Palantir, Jane Street, NASA/JPL, and Stripe.
    • Optional Microsoft SDL review overlay.
    • Profile selection from your request or .parrot/profile.json.
    • A structured review contract that distinguishes blocking issues from advisory observations.
    • MIT license and source attribution.

    Important limitations

    This is an instruction-based review skill, not the full Parrot plugin. It does not include hook-enforced permissions, the builder/checker loop, integrity hashing, or deterministic stop rules. It instructs the agent not to edit files, but does not enforce a sandbox. It does not run tests, and PASS is not a guarantee of correctness or merge readiness. Host-agent compatibility has not been tested across every platform.

    Profile names describe engineering-practice inspiration, not affiliation or endorsement.

    Full plugin and source: https://github.com/devakchow/parrot Adapted from v2.2.0, commit 4403b1054ee39f701d9ebb19e89d37a0940f0711.

    Read-only analysis only. No automated fixes, test execution, or dependency installation. PASS verdict is advisory and does not prove code safety or build success.

    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 any payment.

    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

    2 uses

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Permissions

    Read Files
    Terminal / Shell

    Read repository files and inspect diffs using read-only Git commands. No project-code execution, writes, network access, credentials, or connected accounts are required. Read-only behavior is instructed, not hook-enforced.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    It supplies a repeatable review workflow, profile rubrics, and a structured report that separates blocking findings from advisory observations. Results still depend on the host model and available context; it does not guarantee better results than every generic prompt.

    The instructions are not tied to one language or framework. Review quality depends on the host model and supplied code context; compatibility and quality have not been tested across every language or agent.

    It needs a supplied diff or read access to relevant repository files. Git shell access is optional for inspecting local changes. The skill instructs the agent not to edit, commit, or push, but does not enforce tool permissions or a sandbox.

    The free ZIP includes SKILL.md, six engineering-profile review rubrics, an optional Microsoft SDL review overlay, and the MIT license. It does not include the full Parrot plugin or its hooks.

    No. PASS means no blocking issue was identified within the inspected scope and available context. The skill does not execute tests or prove correctness, security, or merge readiness.

    Load the SKILL.md package with its bundled references into a compatible agent, then provide a diff or repository scope and optionally choose a profile. Installation details depend on the host agent.

    New in Code Quality & Review