majiayu000/spellbook

AGENTS.md Scaffold generates or updates repository-specific AGENTS.md files from repo evidence. It is used when a repo needs scoped agent instructions, validation commands, or directory-specific rules.

API Design covers REST, GraphQL, and gRPC best practices for defining contracts, versioning, pagination, errors, and idempotency. Use it when designing public or internal APIs and their schemas.

Agents Md Optimize audits a CLAUDE.md or AGENTS.md file, scores common instruction patterns, and applies approved fixes in place. Use it when you want to improve an agent instruction file, not just review it.

App User Story QA turns a broad request to test an app into a tracked loop of feature inventory, code-backed expected behavior, and user-story tests. Use it to audit every surface, classify failures, and retest documented defects after fixes.

Architecture Foundation turns architecture discussions into explicit boundaries, contracts, validation gates, and a small execution plan. Use it when designing or refactoring system structure, choosing module or service boundaries, or writing a spec before coding.

Architecture Research helps analyze real systems from source, docs, and runtime evidence before choosing a technical direction. Use it for architecture studies, system archaeology, build-vs-buy decisions, and reassessing earlier choices.

Ask OpenCLI sends prompts to Grok or Gemini through the user's logged-in Chrome session. Use it when you want a second opinion without API costs, or when a workflow needs browser-based access to those models.

Auth Security covers OAuth 2.1 and JWT best practices for implementing login, API authorization, and token handling. Use it for PKCE flows, token rotation, secure storage, and JWT verification.

Brainstorming turns rough ideas into clearer designs through sequential questions and trade-off analysis. Use it when exploring features, architecture, or other open-ended decisions, not for mechanical implementation tasks.

Build Product Demo plans, produces, or diagnoses evidence-backed product demo videos and screen-recorded walkthroughs. Use it when you need a polished promo, launch video, or feature showcase that proves real product behavior.

CLIProxyAPI NewAPI Metering Stack adds a NewAPI billing, rate-limit, and token layer on top of an already verified CLIProxyAPI upstream. Use it to deploy the NewAPI container, bridge container networking, set model pricing, hot-load OAuth credentials, and verify quota accounting.

Capability Distill extracts non-obvious judgment rules from approved strong-model interviews or real task traces into an auditable judgment packet. Use it when you want to preserve key decisions, not when you are just writing process docs or a SKILL.md from scratch.

Clash Doctor inspects Clash Verge and mihomo, manages profiles and routing, and diagnoses connectivity issues. Use it for proxy failures, AI tool routing, local hub alignment, TUN bypasses, or when a sync did not apply.

Clash Routes查看本机 Mihomo 活跃连接，显示进程命中的规则、策略组和真实出口。用于判断 Claude、Codex、浏览器等进程当前走哪条代理链；只读，不修改配置。

Codebase Audit performs adaptive, read-only audits of a project to find design issues, hidden bugs, data-flow breaks, security risks, and technical debt. Use it for full or quick health checks with structured findings and a severity-sorted repair roadmap.

Codex runs OpenAI Codex CLI for code analysis, refactoring, and automated local edits. Use it when you want to inspect a codebase, make workspace changes, or resume an existing Codex session.

Codex Agent Collaboration uses Codex CLI for second-opinion code reviews, cross-verification, debugging help, and alternative implementation ideas. Use it when you want another coding agent to inspect changes before you apply them.

Codex Fluent inspects and maintains Codex local state to keep sessions responsive and low-friction. It helps when chats, logs, or worktrees have grown stale and you need safe archiving with handoffs first.

Codex Log Guard diagnoses Codex local SQLite diagnostic log writes from file, schema, and process evidence. Use it to verify whether logs are protected, actively writing, or safe to clean up or restore.

Codex Retrospective reviews recent session history to find repeated friction and propose minimal AGENTS.md updates or tiny reusable skills. Use it after recurring mistakes or on a weekly or monthly retrospective.

Commit And Push Everything stages all changes, checks for secrets and other risky files, then prompts before committing and pushing. Use it when you want to bundle current work into one commit and send it to the remote.

Local Binary Reverse performs read-only static analysis of user-authorized local binaries and packaged apps. Use it to inspect strings, symbols, dependencies, file types, and version differences without executing the target.

Mobile App UI Design helps design iOS and Android app interfaces, design systems, and accessible layouts. Use it when following Human Interface Guidelines, Material Design 3, or WCAG 2.2 AA requirements.

Test-Driven Development enforces writing a failing test before implementation, then using the red-green-refactor cycle to build and change code safely. Use it for new features, bug fixes, and refactors.

Weekly Report generates an Om Generator weekly report for a given date range by combining Git commits, Claude Code sessions, Codex sessions, and optional local memory records. It is used when a user asks for a Chinese weekly summary document.