Hook Developer
Build Claude Code hooks with correct schemas, registration, and testing patterns
What You Can Do
Create, debug, and test Claude Code hooks with confidence using comprehensive schema documentation and registration patterns. You'll understand the lifecycle of each hook type, their blocking capabilities, and how to structure inputs and outputs correctly. This skill covers all hook types from PreToolUse to SubagentStop, helping you implement the right hook for your automation use case.
Features
understand when each hook fires, blocking capabilities, and primary use cases
precise JSON structures for PreToolUse, PostToolUse, UserPromptSubmit, PermissionRequest, and all event hooks
configure hooks in settings.json with correct syntax and hook type options (command vs prompt)
learn which hooks can block execution and how to implement permission decisions or tool modifications
validate hook inputs/outputs and diagnose schema mismatches
use hooks to load context, set environment variables, and coordinate subagents
auto-approve, deny, or modify permissions based on tool execution context
Example Output
PreToolUse Hook Example:
{
"hookSpecificOutput": {
"hookEventName": "PreToolUse",
"permissionDecision": "allow",
"modifiedToolInput": {
"file_path": "/safe/path/file.txt",
"command": "cat"
}
},
"output": "Tool execution approved and sanitized"
}
UserPromptSubmit Hook Example:
{
"hookSpecificOutput": {
"hookEventName": "UserPromptSubmit",
"permissionDecision": "allow",
"injectedSystemContext": "User has elevated permissions. Session started at 2024-01-15."
},
"output": "Prompt accepted with injected context"
}
SessionStart Hook Example:
{
"hookSpecificOutput": {
"hookEventName": "SessionStart",
"environmentVariables": {
"PROJECT_ROOT": "/workspace",
"API_KEY": "***"
}
},
"output": "Environment initialized"
}
What's Included
- hook-developer SKILL.md: comprehensive reference with all hook types and schemas
- Hook lifecycle table: quick reference showing which hooks can block, inject context, or react to events
- Input/output schema templates: JSON structures for each hook type (PreToolUse, PostToolUse, UserPromptSubmit, PermissionRequest, SessionStart, SessionEnd, Stop, SubagentStart, SubagentStop, PreCompact, Notification)
- Registration checklist: settings.json configuration patterns and hook type options (command vs prompt)
- Testing patterns: validation scripts and debugging workflows for hook execution
Who It's For
- DevOps engineers building automation hooks for Claude Code
- Platform engineers implementing permission and context management systems
- Backend developers integrating Claude into multi-agent architectures
- AI/ML engineers coordinating subagent execution and lifecycle events
- Systems architects designing hook-based approval and validation workflows
Best For
- Blocking or modifying tool execution before it runs
- Validating user prompts and injecting context dynamically
- Handling permission requests and auto-approval workflows
- Coordinating session lifecycle (startup cleanup, environment setup)
- Managing subagent execution and forcing continuation logic
- Debugging hook input/output schema mismatches
- Building notification and alerting systems
You might also like
This skill maps construction claims as interdependent probability states rather than isolated analyses, allowing you to evaluate how changes in one contractual interpretation cascade through damages calculations, delay causation, and liability allocation. You can simultaneously model competing scenarios—concurrent causation, ambiguous change order triggers, insurance coverage dependencies—and identify which settlement positions minimize exposure across all correlated outcomes, not just single interpretations.
by parcadei
You can create, list, enable, disable, and run markdown-based AI automation pipes that execute on schedules ranging from every 30 minutes to custom cron expressions. Manage service connections to Telegram, Slack, Discord, and other platforms directly from the CLI, and troubleshoot pipe execution with built-in logging and debugging commands.
by screenpipe
You'll follow the strict TDD workflow: write a failing test that defines expected behavior, verify it fails, write minimal production code to pass the test, then refactor for quality. This approach ensures complete test coverage, catches bugs early, and produces well-designed code backed by comprehensive test suites. Claude guides you through each phase with examples and enforces TDD principles to prevent shortcuts.
by parcadei
You can formally verify mathematical proofs by translating theorems into Lean 4 code, leveraging Mathlib's extensive library of formalized mathematics. The skill guides you through researching existing lemmas, designing proof strategies, testing approaches, implementing formal proofs, and verifying correctness—turning informal mathematical claims into machine-checked theorems without requiring deep Lean syntax expertise.
by parcadei
You can rapidly process multiple subcontractor bid responses—from trade contractors, material suppliers, and specialty vendors—into a cohesive, auditable estimate structure. Claude extracts key bid components, identifies pricing inconsistencies against benchmarks or prior projects, flags scope gaps and exclusions, and organizes everything by CSI division for seamless estimate assembly. This eliminates manual spreadsheet comparison and reduces the risk of overlooked scope omissions or pricing anomalies when managing 15+ concurrent bids.
by parcadei
You can deploy this agent to execute specific implementation tasks within a larger development plan. It operates independently with fresh context, follows test-driven development principles for every code change, and creates comprehensive handoff documentation that captures progress, decisions, and learnings for the next task executor. This ensures consistent code quality, prevents rework, and maintains continuity across distributed implementation efforts.
by parcadei
Feed your logs, metrics, and traces into Claude to automatically detect anomalies, correlate alerts, and identify root causes in distributed systems. This skill generates step-by-step remediation procedures, prioritizes incidents by severity, and creates incident reports—compressing hours of manual triage into minutes.
You can systematically analyze construction estimates by breaking down components, validating labor and material pricing against market data, identifying embedded errors or outdated assumptions, and applying data-driven contingency allocation. Claude helps you produce risk-adjusted cost projections that are defensible to stakeholders and competitive in bid scenarios, reducing estimate variance and catching pricing gaps before client submission.
by parcadei