Agent Claude Code
Use when cleaning up code after a feature or fix — simplifies for clarity and maintainability while preserving all functionality. Targets recently modified code by default.
Agent Claude Code
Use when querying data, diagnosing database performance, optimizing schemas or indexes, managing backups, replication, monitoring, permissions, or running a database health assessment.
Agent Claude Code
Use when investigating issues, diagnosing errors, examining logs or CI/CD pipelines, debugging test failures, or finding performance bottlenecks — produces a root-cause report.
Agent Claude Code
Use when managing technical documentation — detect docs impacted by code changes, update project and feature docs, and keep docs synchronized with code. Drives /docs-manager --mode=update.
Agent Claude Code
Use when generating E2E tests from recordings or specs, updating visual baselines, or maintaining test-to-spec traceability. Follows the project's configured runner, test format, and evidence contract.
Agent Claude Code
Use when creating, editing, auditing, or refactoring the portable .claude framework itself — skills, agents, workflows, hooks, project-config, framework docs, Codex mirrors. NOT for application code.
Agent Claude Code
Use when creating or modifying frontend/UI components, forms, state, requests, or styling — follow the project's detected platform and documented conventions, using shared abstractions when they exist and fit.
Agent Claude Code
Use when executing a specific implementation phase from a /plan, across backend and frontend, with strict file-ownership boundaries for parallel execution.
Agent Claude Code needs its repo
Use when the user explicitly requests stage, commit, push, or PR creation. Requires operation, scope, and sourceRequest; implementation or review approval never grants Git authority.
Agent Claude Code
Use when generating integration tests, converting test specs into test code, or reviewing integration test traceability. Long-running.
Agent Claude Code
Use when a significant technical difficulty occurs — repeated test failures, production bugs, flawed approaches needing redesign, blocking dependencies, security findings, failed migrations, broken pipelines.
Agent Claude Code
Use when doing web research, knowledge synthesis, course material creation, marketing analysis, or business evaluation — produces structured reports.
Agent Claude Code
Use when investigating slow queries, N+1 patterns, memory leaks, API latency, bundle size, lazy loading, or render performance — adapts to the project's detected stack.
Agent Claude Code
Use when researching and creating an implementation plan for a feature, architecture, or complex solution — before significant implementation or when weighing trade-offs.
Agent Claude Code
Use when researching software development topics — new technologies, documentation, best practices, packages, or open source projects — synthesized into a detailed report.
Agent Claude Code
Use when reviewing authentication, authorization, secret management, input validation, dependency vulnerabilities, OWASP compliance, or service boundary security. Read-only, evidence-backed findings.
Agent Claude Code
Use when planning a new project from scratch with no existing codebase — market research, tech stack evaluation, DDD domain modeling, and staged planning with user validation.
Agent Claude Code
Use when verifying an implementation matches its spec — nothing more, nothing less. Catches spec drift, missing requirements, and extra features. Dispatch BEFORE code-reviewer.
Agent Claude Code
Use when validating code through testing after a feature or significant change — runs unit and integration tests, analyzes results, checks coverage, verifies builds.
Agent Claude Code
Use when doing UI/UX design work — interface designs, wireframes, design systems, user research, responsive layouts, animations — or reviewing new UI for accessibility and responsiveness.
Skill Claude Code needs its repo
Goal: Refresh the portable project AI-context lifecycle — generate or update the Claude root context from project-config.json + template, preserve project instructions through smart merge/refactor, and refresh Codex projections.
Skill Claude Code needs its repo
Goal: Review a plan or a change that calls a model — LLM calls, prompts, agents, tool use, RAG, MCP, evals, guardrails, fine-tuned or classical ML — against the AI-engineering protocol, and report evidence-backed findings ranked by real consequence. Skip when nothing in scope touches a model.
Skill Claude CodeCodex
Goal: Run the selected architecture contract to design a solution, review compliance, grade scalability, or deliver a whole-project health report with evidence-backed decisions.
Skill Claude CodeCodex
Goal: Find real defects in the requested diff and, when requested, converge its fixes without losing coverage or quality.
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: