More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Pirl Review Trial

    1

    Source-reasoned review of an implementation against its plan's quality gates.

    Free

    Security scanned
    pirl-review-trial

    pirl-review-trial

    Example session with this skill installed

    Review the implementation of task T01 (User Auth) in the current diff against the quality gates in plan-v1.md.

    • Read your context and instructions
    • Compiled the pirl-review-trial

    I have completed the gated review for task T01 (User Auth). I analyzed the changes in src/auth/ against the three quality gates defined in plan-v1.md using the trace and architecture lenses.

    Plan-Gated Review Report: T01

    Verdict: FAIL
    Status: REPAIR REQUIRED
    Report Path: reports/PGR-T01-auth-review.md

    IDQuality GateResultEvidence / Finding
    QG.1JWT tokens must expire in < 1 hourPASSconfig.ts line 42 sets expiresIn: '30m'.
    QG.2Passwords must be hashed with Argon2idFAILauth-service.ts uses bcrypt. This violates the architecture constraint.
    QG.3Failed attempts must be logged to auditBLOCKThe Logger injection is missing from the LoginController constructor.

    Findings & Risks

    • FIND-01: Incorrect hashing algorithm. Implementation uses Bcrypt; plan requires Argon2id. (Owner: @dev)
    • BLOCK-01: Missing dependency. AuditLogger is not initialized, preventing gate verification. (Owner: @arch)
    • RISK-01: The session secret is currently hardcoded in env.example. (Accepted trade-off for dev environment).

    Next steps

    1. Switch the hashing implementation in auth-service.ts from bcrypt to argon2.
    2. Inject the AuditLogger into LoginController to satisfy the logging requirement.
    3. Trigger a re-review once the REPAIR disposition is addressed.

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

    About this skill

    Code reviews often drift into bike-shedding or miss the original requirements entirely. This skill enforces a rigorous, plan-based review process that holds implementation accountable to pre-defined quality gates.

    What it does

    • Gate verification: Evaluates every implementation task against its specific quality gates and acceptance criteria.
    • Source-based reasoning: Analyzes diffs and source code to find evidence of completion or failure without needing a runtime environment.
    • Multi-lens analysis: Applies specialized perspectives for trace analysis, adversarial edge cases, requirements, and architecture.
    • Structured reporting: Generates standardized reports classifying issues as findings, risks, or blockers with clear repair instructions.
    • Disposition tracking: Assigns clear next steps like repair, replan, or escalate to ensure the development cycle continues efficiently.

    How it works

    1. Context assembly: Identify the specific plan task, its requirements, and the implementation diff or commit range.
    2. Lens assignment: Map each quality gate to a specialized reviewer lens (e.g., architecture or requirements) for focused analysis.
    3. Evidence gathering: Trace execution paths and verify claims by reasoning over the code and checking for counterexamples.
    4. Classification: Sort identified issues into actionable categories and assign a final verdict for the task.
    5. Report generation: Produce a formal report using a strict template and hand off to the next owner.

    Frameworks & tools

    This skill is language-agnostic and works with any codebase where tasks are defined by requirements and quality gates. It integrates with Git workflows by analyzing commits and diffs.

    Why this beats prompting it yourself

    Generic prompts often produce subjective feedback or miss subtle requirement violations. This skill uses a formal "Review Contract" and specialized lenses to ensure every gate is systematically checked with evidence, preventing reviewers from ignoring hard requirements in favor of stylistic preferences.

    Use cases

    • Validating a Pull Request against the original Jira or GitHub Issue acceptance criteria.
    • Reviewing intermediate tasks in a multi-step implementation plan before proceeding.
    • Verifying bug fixes to ensure the repair actually addresses the root cause without regressions.
    • Audit-trailing quality decisions by recording accepted trade-offs and risks in a persistent report.

    Known limitations

    This skill performs static analysis through reasoning only. It does not execute code, run builds, or trigger automated test suites.

    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.

    Does not execute code or run builds. Reasoning is limited to static analysis of diffs and source. Requires pre-defined quality gates in a plan or ticket.

    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 13 days ago

    • Free with an Agensi account

    Software
    Git

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    File Scopes

    reviews/**
    **/*

    Read Files: the review reads the plan, the diff and the surrounding source, tests and configuration of the repository under review. Everything except the report is read-only. Write Files: the only file written is the review report (default: reviews/**). Product code, tests and the plan are never modified. Terminal / Shell: used for git diff, git log, git show and git status, and for searching the code. No build, test run, app start or script is executed as evidence; another command runs only when the user explicitly asks for it. No network access, browser, environment variables or external accounts are needed. If the user asks to review a remote PR that is not checked out, the agent's own git fetch is the only network use.

    Listed13 days ago

    What's inside

    Frequently Asked Questions

    Unlike standard LLM reviews that focus on style or syntax, this skill uses a "Review Contract" to hold code accountable to your original plan's quality gates, preventing requirement drift and 'bike-shedding.'

    This skill is language-agnostic and functions by analyzing Git diffs and commit ranges against text-based requirements, making it compatible with any programming language or framework.

    No, this skill performs deep reasoning through static analysis of the source code and does not require a runtime environment, build tools, or sandbox execution.

    The skill generates a structured report that classifies findings as blockers, risks, or simple observations, providing clear repair instructions and a final 'Pass/Fail/Replan' verdict.

    The skill is designed to integrate into standard Git-based SDLC workflows, specifically for validating Pull Requests or auditing intermediate steps in complex, multi-task implementation plans.

    You receive the full skill package, including the multi-lens analysis logic and the standardized reporting templates, along with lifetime access to any future version updates.

    New in Code Quality & Review

    Free