AI coding skill for serious reviews

Full-stack quality audits without the hand-wavy parts.

A reusable skill for read-only audits, architecture guidance, project memory, parallel subagents, and CI quality gates across modern software stacks.

Install the skill Explore coverage
Read-only by default Parallel auditors Markdown + JSON output Codex, Cursor, Claude, Antigravity
parallel read-only audit GraphQL skipped
database
done
backend
done
frontend
done
security
done
delivery
done 
review.md
findings/
  backend.findings.json
  security.findings.json
  frontend.findings.json
docs/audit/inventory.json

One skill, many review lenses.

The skill loads only the relevant references and auditors, so a frontend review stays focused while a whole-project audit fans out across the stack.

DatabaseBackendFrontendMobile HTTP APIGraphQLSecurityPrivacy Accessibilityi18nAnalyticsBackground jobs AI/LLM safetyInfrastructureDocumentation Multi-tenancyPaymentsNotifications Import/exportAPI compatibilityDeveloper experience TestingPerformanceObservabilityDelivery
R
Read-only audits

Review mode produces evidence-backed findings without modifying source, schemas, configs, docs, generated files, or lockfiles.

P
Parallel subagents

Whole-project audits can launch one focused auditor per relevant area, then merge and deduplicate the results.

M
Project memory

Optional docs, ADRs, migration backlog, health score, and drift reports help future agents avoid stale assumptions.

Install as a project tool.

Add it as a submodule and point your agent rule file at SKILL.md. The same repository supports Codex, Claude, Cursor, Antigravity, Gemini, and Windsurf. 

git submodule add https://github.com/lablnet/full-stack-quality-skill.git .cursor/tools/full-stack-quality

mkdir -p .cursor/rules

# .cursor/rules/full-stack-quality.mdc
Read .cursor/tools/full-stack-quality/SKILL.md and follow it.

Project context mode for long-running codebases.

When a review needs to become durable project knowledge, the skill can produce agent-facing docs, ADRs, inventory, health trend, migration backlog, and CI gate reports.

  • Audit repo: run relevant read-only auditors and persist per-area JSON findings.
  • Decide target architecture: propose realistic options and record accepted ADRs.
  • Generate docs: create concise guides future agents read before editing.
  • Verify drift: check generated docs against current code and record mismatches.
  • Gate CI: warn or fail PRs that violate architecture, security, tests, or docs.