verify-standards

name	verify-standards
description	Project quality governance scorecard — verifies coding standards compliance across three enforcement tiers: CI gates, AI assistant rules (CLAUDE.md), and team governance (CODEOWNERS, RFC process). Produces a scored checklist with the next recommended action. Adapt the check list to your project's quality standards. Use when: "verify standards", "quality check", "standards audit", "check compliance", "how healthy is the codebase?". Proactively suggest before a major push, at the start of a long session, or after a significant refactor.
origin	generic

Project Quality Standards — Governance Scorecard

Follow every step. Use real tool calls — never assume file state. Customize the checklists below for your project before using this skill.

---

STEP 0 — Context

PROJECT_ROOT=$(git rev-parse --show-toplevel)
echo "Project: $(basename $PROJECT_ROOT)"
echo "Branch:  $(git branch --show-current)"
echo "Commits: $(git rev-list --count HEAD) total"

---

TIER 1 — CI Gates (automated enforcement)

These checks run in CI on every PR. Verify they are configured and passing locally.

# Rust: zero warnings
cargo clippy --lib --no-deps -- -D warnings 2>&1 | tail -5

# Rust: all tests pass
cargo test --lib 2>&1 | tail -5

# C++ (if applicable): build with zero new warnings
# cmake --build build/ --config Release 2>&1 | grep -E "warning:|error:" | head -10

# LOC gate: no file exceeds 1500 lines (adjust threshold to your project)
find src -name "*.cpp" -o -name "*.h" -o -name "*.rs" 2>/dev/null | while read F; do
  LINES=$(wc -l < "$F")
  [ "$LINES" -gt 1500 ] && echo "OVERSIZED: $F ($LINES lines)"
done

Mark each check: ✅ PASS · ⚠️ WARN · ❌ FAIL

---

TIER 2 — AI Assistant Rules (AGENTS.md + CLAUDE.md / project instructions)

These are rules the AI assistant enforces proactively during development.

2a — Cross-tool portability (AGENTS.md)

# AGENTS.md at project root — cross-tool rules (Claude Code, Cursor, Codex, Copilot)
[ -f AGENTS.md ] && echo "AGENTS.md OK ($(wc -l < AGENTS.md) lines)" || echo "MISSING — run: cp /path/to/ai-native-dev-stack/AGENTS.md ."

# CLAUDE.md references AGENTS.md via @include
[ -f CLAUDE.md ] && grep -q "@AGENTS.md" CLAUDE.md \
  && echo "@AGENTS.md present in CLAUDE.md" \
  || echo "CLAUDE.md does not reference @AGENTS.md — add '@AGENTS.md' near the top"

• AGENTS.md present at project root (cross-tool portability)
• CLAUDE.md includes @AGENTS.md so rules load in Claude Code automatically

2b — Project-specific rules (CLAUDE.md)

# CLAUDE.md exists and is substantial
[ -f CLAUDE.md ] && echo "EXISTS ($(wc -l < CLAUDE.md) lines)" || echo "MISSING"

# Check for key sections (adapt to your project)
grep -c "Thread\|Real.time\|Forbidden\|Constraint" CLAUDE.md 2>/dev/null && echo "RT rules present" || echo "RT rules missing"
grep -c "naming\|convention\|pattern" CLAUDE.md 2>/dev/null && echo "Conventions present" || echo "Missing"
grep -c "pre.commit\|build\|test" CLAUDE.md 2>/dev/null && echo "Build rules present" || echo "Missing"
grep -c "error\|unwrap\|Result\|catch" CLAUDE.md 2>/dev/null && echo "Error handling present" || echo "Missing"

Key areas to verify are documented (customize for your project):

• Naming conventions (language, style, abbreviations)
• Thread model and real-time constraints (if applicable)
• Forbidden patterns (globals, singletons, unsafe alloc in hot paths)
• Build and pre-commit workflow
• Error handling policy (unwrap() rules, exception policy)
• File size thresholds and LOC budgets

---

TIER 3 — Team Governance (human process)

These require human action to set up and maintain.

# Architecture Decision Records
ADR_COUNT=$(ls docs/adr/*.md 2>/dev/null | grep -v README | wc -l)
echo "ADRs: $ADR_COUNT"

# CONTRIBUTING.md
[ -f CONTRIBUTING.md ] && echo "CONTRIBUTING.md OK" || echo "MISSING"

# CODEOWNERS (GitHub ownership map)
[ -f .github/CODEOWNERS ] && echo "CODEOWNERS OK" || echo "MISSING (recommended for teams)"

# RFC process
[ -d docs/rfcs ] && echo "RFC dir OK ($(ls docs/rfcs/*.md 2>/dev/null | wc -l) RFCs)" || echo "No RFC dir"

# CHANGELOG
[ -f CHANGELOG.md ] || [ -f CHANGELOG ] && echo "CHANGELOG OK" || echo "MISSING"

---

STEP 4 — Scorecard

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  QUALITY STANDARDS — SCORECARD
  Project: <name> · Branch: <branch>
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Tier 1 — CI Gates            N/M  [results]
Tier 2 — CLAUDE.md Rules     N/M  [results]
Tier 3 — Team Governance     N/M  [results]

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  SCORE: X/Y  |  PASS: A  WARN: B  FAIL: C
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

---

STEP 5 — Next recommended action

Based on the results, recommend ONE concrete next action:

• If any Tier 1 CI check fails → fix it now (stop-the-line rule)
• If CLAUDE.md is missing key sections → add them before next session
• If ADR count is 0 → write ADR-0001 for the most recent architectural decision
• If CODEOWNERS is missing on a team project → create it (15 min, high value)
• If all green → "Standards compliant. No action needed."

---

Customization guide

This skill is a template. Before using it on your project:

1. Replace Tier 1 commands with your actual CI checks (language, tools, thresholds)
2. Replace Tier 2 checklist with your project's CLAUDE.md sections
3. Add project-specific checks (e.g., security scan, coverage gate, performance budget)
4. Adjust LOC threshold (default 1500 — tighten to 800 once codebase is healthy)

Save the customized skill as .claude/skills/verify-standards/SKILL.md in your project.