Workflow Automation

What is Workflow Automation?

Workflow automation means chaining commands, hooks, and agents into pipelines that execute without manual intervention. Instead of running each step yourself, you define the sequence once and let the brain handle it every time.

Think of it like a factory assembly line: raw input goes in one end, and finished output comes out the other. Each station (command, hook, or agent) does its job and passes the result forward.

Designing Automated Pipelines

The most valuable pipelines chain multiple steps into a single trigger. A code review pipeline might look like this:

1. Code Agent writes the implementation based on your spec
2. Review Agent audits for bugs, edge cases, and security issues
3. Test Agent runs or generates tests to verify correctness
4. Deploy step commits and pushes if all checks pass

Each step only runs if the previous step succeeds. This is the same pattern used by CI/CD systems like GitHub Actions, but running locally inside your AI brain.

The /start, /sync, /wrap-up Ritual

Daily workflow automation starts with three built-in commands:

• /start — Loads memory, checks the task board, creates a daily note, and briefs you on priorities
• /sync — Mid-day refresh that processes the scratchpad, updates memory, and re-prioritizes tasks
• /wrap-up — End-of-day routine that logs progress, clears the scratchpad, and updates project status

These rituals ensure context is never lost between sessions. Run them consistently and your brain stays sharp across days and weeks.

Event-Driven Automation

Hooks let you trigger actions automatically when specific events occur:

• Pre-edit hooks — Run validation before any file is modified
• Post-edit hooks — Auto-format, lint, or test after changes are saved
• Commit hooks — Enforce commit message standards and run pre-commit checks
• Session hooks — Save state before compaction, restore context on startup

Event-driven automation eliminates entire categories of human error. You never forget to lint because the hook does it for you.

Error Recovery in Pipelines

Pipelines fail. What matters is how they recover. Good pipeline design includes:

• Retry logic — Transient failures (network, rate limits) get retried automatically
• Fallback steps — If the primary approach fails, try an alternative
• Human escalation — When automated recovery fails, surface the issue clearly to a human
• State preservation — Save progress so you can resume from the last successful step, not from scratch

Practical Exercise

Build a 3-step automated workflow:

1. Create a custom command that generates a new component from a template
2. Add a post-edit hook that runs the linter automatically after the file is created
3. Chain a review agent that audits the generated component for accessibility and TypeScript issues
4. Test the full pipeline end-to-end and verify each step triggers the next