What's new
Deeper diagnostics
Each of the four skill guides has been substantially deepened:
PR Review — analysis now adapts to PR size (scope calibration table); Dependency Disorder and Domain Model Distortion are scanned as independent steps; severity assignments now reference per-risk guides with an explicit tiebreaker rule.
Architecture Audit — new Step 5 scans for missing or collapsed testability seams (places where behavior can be altered without editing source code). Based on Feathers's Seam Model from Working Effectively with Legacy Code. Conway's Law check now includes calibration examples for all three severity tiers.
Tech Debt Assessment — Pain and Spread scores now have concrete calibration examples. New Step 2b classifies each finding as intentional vs accidental debt (Ward Cunningham's original definition). Debt Summary Table gains an Intent column.
Test Quality Review — Test Brittleness and Mock Abuse are now scanned as separate steps with per-step severity thresholds. Characterization Test template added with inline usage guidance. Test suite performance structured into three tiers (>10 min, >30 min, unknown).
More accurate triggering
All four skill trigger descriptions rewritten in natural language and given explicit "Do NOT trigger for:" boundaries. This fixes known false-trigger cases:
brooks-debtno longer fires on HTTP/healthendpoint questionsbrooks-testno longer fires on "how do I write a test?" questionsbrooks-reviewno longer fires on language syntax or from-scratch code questions
Coverage additions
Two Change Propagation concepts are now explicitly named (previously buried):
- Hyrum's Law — implicit API contracts created by observable behavior
- Orthogonality violation — feature changes rippling into unrelated dimensions
Upgrading
# Claude Code
/plugin install brooks-lint@brooks-lint-marketplace
# Codex CLI — re-run your existing install command
# Gemini CLI
gemini skills reloadNo configuration changes required. Existing .brooks-lint.yaml files are fully compatible.