code-standards-review

Review code changes against repo conventions and engineering standards with a verdict report.

1|Updated Mar 24, 2026
One-click install
npx skills add https://github.com/SgtPooki/skills --skill code-standards-review-sgtpooki
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: code-standards-review
Source: https://github.com/SgtPooki/skills/tree/main/skills/code-standards-review
Command: npx skills add https://github.com/SgtPooki/skills --skill code-standards-review-sgtpooki

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Code reviews often produce inconsistent, subjective feedback or miss whether a change fits the codebase's long-term conventions, architecture boundaries, and public contracts. This Skill produces a structured, evidence-backed standards review of any code change — a GitHub PR, branch, local diff, or pasted diff — with a clear PASS, PASS WITH CONCERNS, or FAIL verdict. ## Core Features & Use Cases - Verdict-first report: Emits PASS / PASS WITH CONCERNS / FAIL computed mechanically from findings, in either a full gate mode (with a ten-dimension summary table) or a compact quick mode for local work. - Repo-authority-first review: Discovers AGENTS.md, CLAUDE.md, CONTRIBUTING, ADRs, lint configs, and adjacent exemplars before applying generic principles, so repo conventions override generic SOLID/DRY advice. - Ten review dimensions: Covers repo conventions, architecture boundaries, maintainability, domain model, SOLID/DRY change cost, tests, security, observability, API & semver compatibility, and DX/UX/accessibility. - Greenfield mode: Audits or scaffolds a new project's day-one foundations (CI-enforced formatting, lockfiles, ADRs, AGENTS.md) using the same severity model. - Use Case: Before merging a pull request, ask for a standards review; the Skill inspects the PR via gh, checks the diff against documented repo rules, runs targeted tests, and returns a gate report with file:line-cited findings and a recommended next step. ## Quick Start Review the current branch against main and give me a standards review verdict with evidence-backed findings.

Frequently Asked Questions about code-standards-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review a GitHub PR against repo coding standards?▼

Provide the PR URL or number and request a standards review. The Skill uses gh to fetch PR metadata, the diff, and check status, discovers repo conventions, then emits a gate-mode report with a PASS, PASS WITH CONCERNS, or FAIL verdict and evidence-backed findings.

What is the difference between gate mode and quick mode in a code review?▼

Gate mode produces the full report with a ten-dimension summary table and is used for PRs, pre-merge, and release contexts. Quick mode emits only the verdict, a scope note, findings, and tradeoffs, and is the default for local work and pasted diffs.

Can this review local uncommitted changes instead of a pull request?▼

Yes. It accepts staged, unstaged, or uncommitted local changes, branch comparisons against a merge base, commit ranges, and pasted diffs. It states exactly which changes were included and flags anything that could not be verified.

Does a standards review replace debugging or bug hunting?▼

No. It evaluates whether a change preserves conventions, boundaries, contracts, tests, and maintainability rather than diagnosing why code fails. A correctness defect introduced by the diff is still reported when found, but it is not an exhaustive defect hunt.

When should I use greenfield mode instead of a diff review?▼

Use greenfield mode when bootstrapping a new repository or auditing a young project's foundations rather than reviewing a diff. It checks day-one items like CI-enforced formatting, lockfiles, ADRs, and AGENTS.md, reporting gaps with the same severity model.

Why did my review return PASS WITH CONCERNS instead of PASS?▼

PASS WITH CONCERNS means there are no blocker findings but at least one should-fix finding, an unresolved tradeoff, or a relevant dimension marked not assessed. The verdict is computed mechanically from findings, so the report always lists what caused it.