reviewing

Detect defects against the plan, architecture, code, and policies; classify findings without proposing improvements.

⭐0stars🔀0forks📁View on GitHub🕐Updated Jan 10, 2026⬇️2downloads

When & Why to Use This Skill

This Claude skill provides an automated, objective auditing mechanism for software development artifacts. It systematically detects defects within codebases, architectural designs, and project plans by cross-referencing them with predefined policies and requirements. By classifying findings into severity levels—Blocking, Fixable, or Advisory—it delivers a clear impact assessment without the distraction of remediation suggestions, ensuring a focused and unbiased quality assurance process.

Use Cases

  • Automated Pull Request Auditing: Scan new code submissions for policy violations and critical defects to ensure high-quality merges and maintain code standards.
  • Architectural Integrity Verification: Evaluate system design documents against organizational standards to identify structural weaknesses or non-compliance early in the development lifecycle.
  • Compliance and Policy Enforcement: Review technical documentation and project plans to ensure strict adherence to internal security protocols and industry regulations.
namereviewing
descriptionDetect defects against the plan, architecture, code, and policies; classify findings without proposing improvements.

Purpose

Report defects with required classifications and clear impact, without remediation.

Inputs

  • Artifacts under review (code, plan, architecture)
  • Applicable policies and context

Steps

  1. Load required context: plan, architecture, policies, and artifacts.
  2. Identify defects against requirements, policies, and correctness.
  3. Classify each finding as Blocking, Fixable, or Advisory.
  4. Report precise locations and impacts using the required format.

Outputs

  • Markdown findings with classification, location, issue, and impact; no fixes or redesigns.

Failure modes

  • Suggesting improvements or scope changes
  • Missing required classifications or unclear locations
  • Mixing detection with remediation