v1.23.0

Added

• /loop integration: workflow.md now has a "Loop & Monitoring" section with patterns for CI watch, incident health monitoring, and Docker builds; /session-wrap reference added to Stopping Protocol
• ci-watch skill + command: single-purpose CI status checker (gh run list), designed to run standalone or via /loop 2m ci-watch
• ci-monitor agent (haiku): lightweight agent that checks GitHub Actions status and reports pass/fail/in-progress with failing step details
• session-wrap skill + command: end-of-session protocol — captures [LEARN] items to patterns.md, updates scratchpad with next steps, checks for uncommitted changes, asks if polaris needs updating
• loop-suggest.js hook: PostToolUse advisory after git push — suggests /loop 2m ci-watch to monitor CI (non-blocking)
• commit skill: Step 9e now suggests /ci-watch or /loop 2m ci-watch after push
• incident-response skill: Phase 2 (Investigation) now suggests /loop 1m check logs for live service monitoring
• All new items registered in sync-config.json for /sync-global propagation

v1.22.0

Added

• UX workflow: /ux 3-phase skill (clarify → wireframe → implement), ux-guard.js advisory hook (reminds to run /ux before editing .tsx/.jsx UI files), ux.md rule (mandatory gates: wireframe approval, mobile-first, async states), ux.md command — full UX enforcement system
• Commit flash audit (Step 0): /commit now detects accumulated [flash] commits before running the quality gate, so the gate covers all changes since the last full commit
• Smart stack detection (Step 5): /commit now inspects staged files to run only the affected stack's tests (avoids running the full multi-stack suite for a single-stack change)
• After-plan commit-flash rule: workflow.md now specifies that after executing a plan, the final action is to run /commit-flash
• Always-recommend convention: conventions.md now mandates stating a recommendation when asking the PO a question
• UX skill, command, hook, and rule added to sync-config.json for future /sync-global propagation

v1.21.0

Added

• "Research before acting" system: rule + hook that forces Claude to check official docs (context7), forums (WebSearch), and community best practices before acting in high-risk situations — package installs, plan mode, error loops, brainstorming, new tech, API calls, and config changes
• research-first.js hook (PreToolUse Bash/Agent + PostToolUse Bash): detects package installs, Plan subagent entry, and consecutive error loops (threshold: 2) with stateful tracking via session temp files
• research-first.md rule: behavioral guidance covering 6 situations where Claude must research before acting, including brainstorm/ideation for collective intelligence

v1.20.1

Fixed

• block-secrets hook now uses ask instead of deny for all sensitive files — no more silent blocks on .env*, config.json, etc.

v1.20.0

Added

• Spec integration across workflow: TDD Step 0 (read spec before RED), DoR/DoD spec requirements, review spec pre-check, advisory spec-reminder hook
• Tech-debt audit expanded to 4 holistic dimensions: Code Health, Test & Reliability, Documentation & Specs, Dependencies & Infra — replacing the previous 6 flat categories
• New debt categories: spec coverage/freshness, config docs, error handling, build/CI health, with security delegating to existing security-auditor agent
• Code review and code-simplifier steps (3b/3c) in /commit workflow for 3+ file changes

v1.19.1

Added

• Plugins reference in rules (plugins.md) listing all installed plugin capabilities, triggers, and agents

v1.19.0

Added

• Multi-select team priorities in /tech-debt-audit: users can now select multiple dimensions (maintainability, reliability, performance, testability) instead of being limited to one

v1.18.0

Added

• SDD (Spec-Driven Development) as principle #7 in development conventions
• Init mode for /spec-update: bootstrap functional specs for all domains on a new project via codebase scan with 4-pass heuristic

v1.17.0

Added

• Check 11 coverage-config in /audit-conformity: audits coverage configuration quality (threshold, branch, source filtering, report output) across Python, Node.js, Go, and Rust stacks

Removed

• auto-format.js PostToolUse hook: removed due to critical bug (formatter emptying files — 5/7 files corrupted in coachbywill)

v1.16.0

Added

• auto-format.js PostToolUse hook: auto-runs project formatter (Prettier/ruff/gofmt/rustfmt) after each Edit/Write
• debug-warning.js PostToolUse hook: warns when edits introduce debug artifacts (console.log, print, etc.)

Fixed

• Corrected deny → ask terminology in audit-conformity, sync-global, and GUIDE.md to match actual permissions model