task-orchestrator-hooks-builder_skill

Overview

This skill helps you design, build, and test hooks that integrate Task Orchestrator cascade events into your workflows. It guides you from requirement interviews through script generation, configuration, and testing so hooks run reliably across any MCP client. The focus is on practical, defensive bash hooks that handle cascade events, flow-aware behavior, and blocking/non-blocking actions.

How this skill works

I walk you through choosing the right trigger event (PostToolUse, SubagentStop, PreToolUse), selecting the tool/matcher to watch, and defining the matcher conditions (operation, containerType, status). I generate a portable bash hook that reads JSON from stdin, performs defensive checks, verifies dependencies, and executes the requested action (git commit, run tests, notify, log metrics, call external APIs). I also provide the exact .claude/settings.local.json snippet, test JSON payloads, and troubleshooting steps for fast validation.

When to use it

• When you want automated reactions to Task Orchestrator events (task/feature/project changes)
• When you need quality gates that can block feature completion until tests pass
• When you want to auto-commit, log metrics, or update external systems on task events
• When you need flow-aware hooks that adapt for prototype/secure/normal flows
• When you want to record or analyze cascade event analytics across workflows

Best practices

• Start with non-blocking logging or metrics before adding blocking behavior
• Always perform defensive input checks and exit 0 on unmatched conditions
• Use $CLAUDE_PROJECT_DIR for paths and verify required tools (jq, sqlite3, git) at script start
• Keep hooks fast; use timeouts in settings.local.json for long work
• Document trigger, matcher, actions and provide sample JSON for testing
• Version control hooks and test locally before enabling in Claude

Example use cases

• Auto-commit task metadata when a task status becomes completed (PostToolUse manage_container, operation=setStatus, containerType=task)
• Quality gate that runs tests and blocks feature completion if tests fail (blocking PostToolUse on feature setStatus=completed)
• Send notifications or webhooks when subagents finish work (SubagentStop event)
• Append task duration metrics to CSV when a task completes for analytics
• Flow-aware cascade hook that skips tests for rapid_prototype_flow but enforces them for with_review_flow