LucasSantana-Dev Config Drift Detect
Project gates that are too strict don't make a project safer — they force workarounds, test bloat, and skill bail-outs. This skill audits every gate and flags incompatibilities with the project's app type, before they cause concrete pain.
Auto-invocation
- Scheduled: weekly via
~/Library/LaunchAgents/com.lucas.config-drift.plist(Mondays 03:30). Per-repo report written to~/.claude/projects/<slug>/memory/config_drift.md. - Triggered automatically by Claude when:
- User asks to refactor, restructure, or "clean up" anything that touches gated code
- Any skill stalls with "the gate prevents me from..." (especially test-cleanup, ship)
- Before
/test-cleanup,/refactor,/shipon a repo this skill hasn't audited recently
Workflow
1. Detect app type and source size
cd "$REPO_PATH"
LOC=$(find src lib app extension -name '*.ts' -o -name '*.js' -o -name '*.py' \
-o -name '*.go' 2>/dev/null | grep -vE "spec|test|node_modules" \
| xargs wc -l 2>/dev/null | tail -1 | awk '{print $1}')
# App-type heuristics from package.json / source structure
APP_TYPE=$(grep -lE "discord\.js|chrome\.runtime|webextension|express|fastify|nestjs|next|cli|commander" \
package.json 2>/dev/null | head -1)
Map to defaults table:
| App type | LOC | Sane coverage gate | Sane test count | Sane suite runtime |
|---|---|---|---|---|
| Discord bot | ~5k | lines ≥80%, fn ≥80% | 50–200 | <30s |
| Browser extension | ~3k | lines ≥80%, fn ≥80% | 40–150 | <20s |
| REST API | ~4k | lines ≥85%, fn ≥85% | 80–250 | <60s |
| CLI tool | ~2k | lines ≥80%, fn ≥80% | 30–120 | <15s |
| Full-stack app | ~15k | lines ≥75%, fn ≥75% | 200–600 | <2min |
| Library (published) | varies | lines ≥90%, fn ≥90% | proportional | <30s |
2. Audit each gate
# Coverage thresholds
jq '.coverageThreshold' jest.config.json package.json 2>/dev/null
grep -A10 "coverageThreshold\|threshold:" jest.config.* vitest.config.* 2>/dev/null
# TypeScript strictness
jq '.compilerOptions | with_entries(select(.key | test("strict|noImplicit|exactOptional")))' tsconfig.json 2>/dev/null
# ESLint rules count + severity distribution
test -f .eslintrc.json && jq '.rules | length' .eslintrc.json
test -f eslint.config.* && grep -cE "'error'|'warn'" eslint.config.* | head -3
# Husky/lint-staged gates
test -f .husky/pre-commit && cat .husky/pre-commit
test -f package.json && jq '."lint-staged"' package.json 2>/dev/null
# SonarCloud quality gate
test -f sonar-project.properties && grep -E "^sonar\." sonar-project.properties
# Branch protection (GitHub)
gh api "repos/:owner/:repo/branches/main/protection" 2>/dev/null \
| jq '{required_status_checks: .required_status_checks.contexts, required_review_count: .required_pull_request_reviews.required_approving_review_count}'
3. Compare each gate to the app-type default
For each gate found, compute:
- delta — current value vs sane default for app type
- implied cost — what does this gate force the team to do?
- compatibility — is this gate compatible with normal proportionality targets?
Example:
Gate: jest coverage functions ≥99%
App type: browser extension
Default for type: ≥80%
Delta: +19pp (significantly stricter)
Implied cost: ~1480 tests minimum (one per function); proportionality target 40-150 unreachable
Compatibility: INCOMPATIBLE with /test-cleanup target
Recommended: lower to 85% (industry norm for browser extensions)
4. Severity classification
| Severity | Trigger |
|---|---|
| CRITICAL | Gate makes a normal workflow impossible (test-cleanup, ship, refactor) |
| HIGH | Gate is >15pp stricter than sane default for app type |
| MEDIUM | Gate is 5-15pp stricter than default |
| INFO | Gate is within ±5pp of default |
| LOOSE | Gate is significantly looser than default (rarely a problem; flag only if security-relevant) |
5. Output
CONFIG DRIFT REPORT — <repo>
App type: Browser extension (~3k LOC)
Audited: 7 gate categories
CRITICAL (1):
jest functions ≥99% on extension/lib/**
→ makes /test-cleanup target (40-150) unreachable
→ Recommend: lower to 85% with /gate-relax (when implemented)
HIGH (2):
tsconfig strict: false (sane default: true for new code)
→ Allows implicit any; lets bugs through static analysis
ESLint warnings: 0 errors, 230 warnings
→ Warnings are systematically ignored; prune or promote
MEDIUM (1):
Husky pre-commit runs full test suite (~5s)
→ Slows commits; recommend lint-staged only
INFO (3):
branch protection, sonar quality gate, prettier — within sane defaults
6. Write to memory
Output to ~/.claude/projects/<slug>/memory/config_drift_<repo>.md with frontmatter so
next session sees it. Include the recommended remediations as actionable bullets.
7. Optional: chain to gate-relax
If /gate-relax skill exists and severity ≥HIGH, propose chaining to apply the fix
automatically.
Outputs / Evidence
- One-page severity-classified gate report per repo
- Recommended remediations with target values
- Memory file for next-session context
Failure / Stop Conditions
- No recognized app type → output "unknown app type, manual review needed"
- Gate config files unreadable → report which ones, continue with the rest
- gh CLI unauthenticated → skip branch protection, continue with rest
Memory Hooks
- Read prior drift reports to surface trend ("function gate raised from 95→99% last month, now incompatible")
- Write the report itself as the primary output