claude-agent-dev

Autonomous code execution agent. Executes tasks, refactors code, and applies improvements to any codebase you're pointed at. Use this agent when you need to: - Implement a specific feature or fix a bug in a codebase. - Refactor or clean up existing code for better maintainability. - Run diagnostic scripts or tests to identify issues. - Perform batch updates or automated changes across multiple files. <example> "Implement the retry logic in hooks/handlers/format.mjs and run the tests to confirm it passes." </example> *Note: This agent requires the `managed-agents-2026-04-01` beta header.*

Code debugging specialist. Diagnoses runtime failures, logic bugs, and quality inconveniences through systematic root-cause analysis so the main thread does not have to context-switch between reading code, logs, and tracing call chains. Use this agent when you need to: - Identify the root cause of a runtime failure or logic bug. - Analyze error logs and stack traces to trace back to the source of an issue. - Classify bugs by severity and category to prioritize fixes. - Propose code fixes as diffs or replacement blocks without applying them. <example> "Investigate the NullPointerException in the user authentication flow and explain the root cause with a proposed fix." </example>

Documentation maintenance agent. Audits, writes, and keeps all docs in sync with the codebase. Use this agent when you need to: - Create or update README files, API docs, or inline code comments. - Audit existing documentation for staleness, gaps, or inaccuracies. - Generate architecture decision records (ADRs) or changelog entries. - Write skill documentation, agent descriptions, or hook reference guides. - Synchronize docs after a refactor — ensuring names, paths, and examples are current. <example> "Update the README to reflect the new hook runner architecture and add a Mermaid diagram." </example> *Note: This agent requires the `managed-agents-2026-04-01` beta header.*

Research-focused agent for exploring codebases and documentation. Use this agent when you need to: - Explore a new codebase and understand its structure and dependencies. - Research documentation for libraries, frameworks, or external APIs. - Search for specific symbols, patterns, or file usages across a project. - Gather information before starting a complex task. <example> "Explore the hook handling logic and find all places where PreToolUse is emitted." </example> *Note: This agent requires the `managed-agents-2026-04-01` beta header.*

Create, audit, refactor, or trim AGENTS.md, CLAUDE.md, GEMINI.md, and onboarding guides. Trigger on 'agent docs', 'instructions', 'onboard me', 'understand this repo', 'setup AGENTS.md', 'this file is too long', 'trim CLAUDE.md', 'improve agent instructions', 'add conventions', 'update instructions file', or when the user shares an existing AGENTS.md/CLAUDE.md and asks to improve, review, or clean it up.

Use when a codebase has structural problems (circular deps, God modules, testability issues) or when designing new systems. Trigger on 'architecture review', 'where should this code live', 'too coupled', 'God class', 'design this system'.

Structured requirements discovery before implementation. Trigger on 'let's build', 'add a feature', 'we need a new', 'I want to implement', 'add X to', 'create a Y', ambiguous design, or unclear terminology — even when the user says 'just build it'. Proactively offer before any implementation begins. Prevents rework by catching problems early.

Mandatory quality gate before delivery. Trigger on 'code review', 'review this diff', 'any issues with this', 'check for bugs', 'quality review', 'ready to merge'. ALSO trigger automatically: (1) after verification-before-completion confirms tests pass, (2) before opening a PR/shipping the change, (3) on any non-trivial change that will be committed. Do not skip — correctness bugs and security issues discovered here are cheaper to fix than after merge.

Design, write, test agents and subagents. Trigger on 'build agent', 'create subagent', 'agent prompt', 'multi-agent', 'managed agent', 'agent not working', 'agent keeps triggering wrong thing'. Also: agent role selection, permission scoping, tool surface design, choosing between subagent/team/workflow/managed agent.

10 hooks across 7 events

An interactive CLI tool for building and validating Claude Code agents, skills, hooks, and commands. Follow the Design → Build → Validate → Ship cycle with phase-aware output.

Hook handler unit tests — live feedback during hook development

Skill script tests — live feedback during skill development

Hook execution telemetry — surfaces errors and slow hooks in real time (PowerShell)