13 Hooks for Automated Safety & Quality

Automatically searches and loads relevant SKILL.md files before any task starts. 61 domain mappings ensure the right skills are loaded for every task type.

Reads memory.md and knowledge-base.md before task execution. Ensures the agent starts every task with full project context and learned patterns.

Routes tasks to the correct specialist agent based on task type. Coder, reviewer, tester, researcher — the right agent handles each job.

Auto-formats code after every edit using Prettier and ESLint. Consistent style across your entire codebase without manual intervention.

Runs TypeScript type checking after file changes. Catches type errors immediately — before they compound into larger issues.

Runs affected tests after code modifications. Validates that changes work and existing functionality remains intact.

Updates memory.md with task results and learnings after every completed task. Persistent context that carries across sessions.

Extracts reusable patterns from completed work and promotes them to knowledge-base.md. The brain gets smarter with every task.

Logs quality scores and metrics per task — correctness, completeness, efficiency, security. Tracks performance over time.

Saves state before auto-compaction triggers. Prevents context loss during long sessions by checkpointing memory automatically.

Prevents accidental commits to the main branch. Forces feature branch workflow to keep production deployments safe.

Blocks commits containing API keys, tokens, or credentials. Scans staged files before every commit to prevent secret leaks.

Detects retry loops and escalates after 3 consecutive failures. Stops the agent from wasting tokens on the same error.