An agent for checking captured traffic from a proxy server. It verifies whether the application correctly detects five kinds of agent-related content, including project instructions, output styles, slash commands, skills, and sub-agents.
A code-review agent examines changes in a pull request or local Git branch. It checks for bugs, security problems, error handling gaps, performance issues, design concerns, tests, and naming.
An agent for diagnosing and fixing CSS and JavaScript problems in the Claude Inspector interface. It follows a reproduce, gather evidence, form hypotheses, and test process.
A macOS release process for Claude Inspector. It covers building signed and notarized application packages, creating a GitHub release, and updating a Homebrew cask.
Build and run an AI development loop — collects pipeline, iterations, checklist, skills via conversation, then generates and executes a staged claude -p loop.
Skill "test-bouncer" from kangraemin/ai-bouncer, covering test-bouncer, 사용 시점, 실행, 1. 전체 e2e (설치 → hook → update → uninstall → 재설치) and 2. hook 단위 테스트 (소스 기준).
A development agent that implements a step assigned by a lead developer. It checks that the build succeeds and reports the result in a required format.
An agent for writing end-to-end tests, which test a complete user flow through an application. Its description also mentions restoring context, collecting scenarios, and following a work sequence.
An orchestrator that executes an approved plan by splitting development into phases and coordinating developers and quality assurance. It checks whether each step meets its completion criteria.
A skill for the bouncer-status command from kangraemin/ai-bouncer. It first reads configuration and then performs a health check, preferring local settings before global ones.