A black-box quality checker that starts an application and tests it from a user's point of view in a real runtime or browser. Black-box testing checks behavior without changing or inspecting the implementation as the main test method.
A codebase research agent that deeply examines folders, files, dependencies, call flows, and business rules, then saves its findings in a research.md file.
A web-research agent that investigates external technical specifications, such as library behavior, APIs, standards, and migration guidance. It writes findings to a web_research.md file under a task-specific documentation folder.
Instructions for leee880619-commits/ClaudeCode-Harness-Setup-Assistant, covering harness-architect (plugin development context), 이 레포가 무엇인가, 이 레포에서 개발 시 (contributors only), 수정 시 참고할 문서 and 중요 원칙 (요약 — 정본은 .claude/rules/.md).
Unified audit entry — runs harness-audit (configuration integrity), ops-audit (runtime debt), and fit-audit (project-fit drift) in parallel and assembles a single consolidated report. This is the only audit entry point exposed in slash-command autocomplete.
Audit the fit between an already-deployed Claude Code harness and the current state of the target project (track drift, archetype/quality-axis misfit, agent scale misfit, permission-path drift, CLAUDE.md/domain identity drift, MCP/hook external-interface drift). Diagnosis only — improvements are delegated to…
Audit an existing Claude Code harness for configuration integrity (4-scope scan, JSON validity, anti-patterns, permission safety, agent↔playbook mapping). Diagnosis only — improvements are delegated to /harness-architect:harness-setup re-run. Complementary to /harness-architect:harness-setup Phase 9 (build-time…
Build a complete Claude Code harness for any project via a 9-phase orchestrated workflow (scan, workflow design, agent team, playbooks, hooks/MCP, validation).
A help command for a tool that builds or audits a Claude Code project setup, including instructions, settings, rules, agents, playbooks, hooks, and MCP connections. MCP is a way for an agent to connect to external tools.
A research agent that searches the web for sources, collects page content, and summarizes the findings in a structured report. It records source links and extraction dates.