Self-evolving skill packs that give Claude Code & Codex judgment, not knowledge — decidable rules, per-playbook regression quizzes for models, and a project harness generator. zh-TW canon + full EN mirror.
Agent Claude Code
Use this agent to perform a thorough, line-level code review of recently written or changed code. This is the FIRST half of the two-agent cross-validation review workflow (paired with system-architect). It focuses on correctness, bugs, security (OWASP Top 10), style/convention adherence, and maintainability at the…
Agent Claude Code
Use this agent for data collection, research, and fact verification tasks. It gathers information from the web and other sources, cross-checks claims against multiple independent sources, cites everything, and clearly separates verified facts from unverified ones. Use it when you need to research a topic, validate a…
Agent Claude Code
Use this agent to debug a specific failing behavior — a bug, crash, error message, wrong output, or unexpected runtime behavior — where the goal is root cause + a verified fix. Distinct from fullstack-code-generator (which builds new features) and test-automation-expert (which writes tests). Examples:\n\n \nContext: A…
Agent Claude Code
The single implementation agent — use it to write or refactor production code on any layer and stack (Flutter/Dart, frontend web React/Next, backend Node/Go, database, infra). Absorbs former Flutter-specific coding; distinct from debug-expert (fixes a specific bug) and test-automation-expert (writes tests).…
Agent Claude Code
Use this agent for design-level judgment on a codebase, in two modes. MODE 1 — the SECOND half of the two-agent cross-validation code review workflow (paired with code-reviewer): evaluates a change's architecture soundness, design patterns, coupling/cohesion, scalability, data flow, API/contract design. MODE 2 …
Agent Claude Code
Use this agent to create, review, or optimize tests — unit, integration, e2e, and CI/CD test strategy — across any stack. Distinct from fullstack-code-generator (writes product code) and debug-expert (fixes a specific bug); this agent writes tests and does not touch product code. Examples:\n\n \nContext: New code…
Agent Claude Code
Use this agent to actually implement or modify UI — build a screen/component, adjust layout, styling, spacing, responsive behavior, or UI copy. This is the hands-on builder; analysis and recommendations are handled by ux-improvement-advisor (they interlock: advice → advisor, doing → this agent). Examples:\n\n…
Agent Claude Code
Use this agent for read-only UX analysis and actionable improvement recommendations on interfaces, user flows, or interaction patterns — heuristic evaluation, usability/accessibility critique, flow friction. It advises but never edits; landing the changes is handled by ui-design-expert (they interlock: analysis → this…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: