fanshi_personal_skills

Guide for assigning tools to agents and configuring `tool_choice` in multi-agent systems. Covers allowlists/blocklists (`tools` vs `disallowedTools`), wildcard expansion, `filterToolsForAgent`, built-in agent scoping, ToolSelector buckets, `tool_choice` modes (auto, any, forced), and agent schemas with `lazySchema` and `BuiltInAgentDefinition`. Use when the user says "agent tool allocation", "tool allocation", "tool_choice", "which tools for which agent", "agent tool scoping", "allowlist blocklist", "filterToolsForAgent", "disallowedTools", "lazySchema", "BuiltInAgentDefinition", "ASYNC_AGENT_ALLOWED_TOOLS", "COORDINATOR_MODE_ALLOWED_TOOLS", or "agent tool boundaries"; also use when designing a new agent's tool access. For tool descriptions use `mcp-tool-design`; for errors use `mcp-error-response`; for MCP optimization use `mcp-tool-enhancement`.

Design input and output JSON Schemas for search or retrieval tools that query the web, databases, code indexes, vector stores, or external services. Use when designing, reviewing, or refactoring a search-result tool, even if the user does not say "schema". Covers required vs optional/nullable fields, empty results, missing metadata, truncation/pagination (`appliedLimit` vs `truncated`), heterogeneous results (`z.union` and discriminated unions), enum `other` vs `unclear`, and HTTP metadata. Trigger on "search tool", "database/web/vector search", "retrieval output schema", "pagination", "truncation", "no results", "nullable fields", "enum other/unclear", or hallucinated missing fields. For general tool description or splitting, use `mcp-tool-design`; for error shapes, use `mcp-error-response`.

Generate or update `.claude/rules/tool-boundary.md` from an approved `docs/tool-routing-report.md`. Use when the developer says "boundary", "load tools", "apply routing", "generate boundaries", or approves a routing report and wants per-tool capability, limitation, chain-pattern, and failure-behavior declarations. Requires a routing report with `status: approved`; if missing or unapproved, stop and tell the user to run `/sentinel:routing` or approve the report first. Do not use for scanning or classifying which skills a project needs; use `/sentinel:routing` for that. Preserve manual edits marked with `<!-- manual -->` when updating an existing boundary file.

所有联网操作必须通过此 skill 处理，包括：搜索、网页抓取、登录后操作、网络交互等。 触发场景：用户要求搜索信息、查看网页内容、访问需要登录的网站、操作网页界面、抓取社交媒体内容（小红书、微博、推特等）、读取动态渲染页面、以及任何需要真实浏览器环境的网络任务。

Guide for locally testing Claude Code CI pipelines with structured JSON output. Assumes CLAUDE_CODE_OAUTH_TOKEN is already configured. Focuses on extracting structured_output from `--json-schema` results using jq, with precheck, output redirection, and CLAUDE.md context loading. Use when the user says "test CI locally", "claude -p", "structured output", "json-schema output", "extract CI results", "jq structured_output", "CI pipeline output", or wants to verify their CI JSON output configuration works before pushing to a real pipeline.

Scaffold a Python FastMCP server for domain-specific authenticated web search (Xiaohongshu, Dianping, Weibo, LinkedIn, NYT, region-locked forums, etc.) where the user's browser session supplies auth. Produces a working server skeleton with four wired-in features: Playwright-MCP browser bridge (with Claude-in-Chrome fallback), EN-to-domain-native query translation, per-search scratchpad directory, and structured writeback with source attribution. ALWAYS use this skill when the user says "build an MCP server for <site>", "scaffold a search MCP", "domain search server", "FastMCP server for <site>", "search scraper MCP", or asks to wrap a site behind an MCP search tool. Do NOT hand-write a FastMCP search server — invoke this skill so attribution, scratchpad layout, and backend selection stay consistent. For generic MCP tool shape guidance, see mcp-tool-design; for search result schema, see search-tool-schema-design.

Guide for designing structured error responses in MCP tools. Covers error classification (transient, validation, business, permission), the isError flag in MCP protocol and Anthropic API, retryable vs non-retryable errors, the critical distinction between access failures and valid empty results, structured error metadata design, and subagent error propagation patterns. Use when the user says "MCP error handling", "error response design", "isError flag", "tool error classification", "retryable errors", "error propagation", "tool returns wrong error", "empty result vs error", "when to retry MCP tool", "subagent error handling", "structured error response", "access failure vs empty result", or wants to design how MCP tools should report and classify errors for LLM agents. For tool description design, see mcp-tool-design instead. For platform-level MCP optimization, see mcp-tool-enhancement instead.

Guide for designing MCP tool content that LLMs can reliably select and use. Covers writing effective tool descriptions, structuring input/output contracts via JSON Schema, splitting tools by single responsibility, avoiding misrouting between similar tools, and coordinating system prompt instructions with tool descriptions. Use when the user says "tool description design", "tool interface design", "MCP tool descriptions", "tool splitting", "input output contract", "tool misrouting", "tool selection", "how to write tool descriptions", or wants to understand how LLMs choose between tools based on description content. For platform-level MCP optimization (searchHint, alwaysLoad, annotations), see mcp-tool-enhancement instead.

Guide for platform-level optimization of MCP server tools within Claude Code's tool pool. Covers discoverability (searchHint, alwaysLoad), permission classification (annotations), MCP Resources as content catalogs, and .mcp.json configuration. Use when tools aren't being selected by the model, when designing MCP Resources, or when configuring MCP server metadata. Also use when the user mentions "MCP tool not being used", "agent ignores my MCP tool", "MCP resource design", "MCP discoverability", "searchHint", "alwaysLoad", "annotations", or ".mcp.json configuration". For tool description design principles and input/output contracts, see mcp-tool-design instead.

Create a new .claude/rules/*.md rule file with proper structure and conditional paths frontmatter when appropriate. ALWAYS use this skill when the user says "create a rule", "add a rule", "new rule", "new rule for...", "write a rule", "add instruction", "create instruction", or wants to add a .claude/rules/ file. This skill guides the agent to determine whether the rule applies globally or only to specific file types/paths, generates the correct frontmatter, and updates README.md afterward. Do NOT create .claude/rules/ files without invoking this skill first.

Build a new skill from idea to working SKILL.md. Orchestrates intent capture, frontmatter design decisions (execution context, tool permissions, arguments), directory scaffolding, and content authoring. Supports quick builds (scaffold + design guidance) and full builds (interview + design + scaffold + test/iterate via skill-creator). Use when the user says "build skill", "create skill", "new skill", "scaffold skill", "make a skill", "I want to build a skill", "add a skill", or wants to go from idea to working skill. For improving or testing an existing skill, invoke skill-creator directly. For agent-level tool boundaries (allowlist/blocklist), see agent-tool-allocation.

AI-driven multi-format SVG content generation system. Converts source documents (PDF/DOCX/URL/Markdown) into high-quality SVG pages and exports to PPTX through multi-role collaboration. Use when user asks to "create PPT", "make presentation", "生成PPT", "做PPT", "制作演示文稿", or mentions "ppt-master".

Write publication-ready ML/AI papers for NeurIPS, ICML, ICLR, ACL, AAAI, COLM. Use when drafting papers from research repos, structuring arguments, verifying citations, or preparing camera-ready submissions. Includes LaTeX templates, reviewer guidelines, and citation verification workflows.

Transform documents into interactive HTML study notes with key concepts, problem-solution pairs, examples, and self-quiz questions. Use when asked to create study notes, summarize lecture slides, or generate learning materials from documents in slides/ folder. Output goes to notes/ folder.

去除文本中的 AI 生成痕迹。适用于编辑或审阅文本，使其听起来更自然、更像人类书写。基于维基百科的"AI 写作特征"综合指南。检测并修复以下模式：夸大的象征意义、宣传性语言、以 -ing 结尾的肤浅分析、模糊的归因、破折号过度使用、三段式法则、AI 词汇、否定式排比、过多的连接性短语。

Guide for packaging local skills as Cowork Plugins and publishing them to a GitHub marketplace for cross-project, cross-device installation and reuse.

Stage, commit, and push all current changes to the remote with a human-readable summary and user approval before executing. Use when the user says "push", "push to github", "push all changes", "commit and push", "check status and push", "send this to remote", or "ship it". Does NOT create branches or pull requests — pushes to the current branch only.

Bootstrap a new project with CLAUDE.md and PROGRESS.md. Use when the user runs /start, says "initialize project", "set up claude.md", or wants to create the external memory system for a new or existing project. Handles both existing projects (auto-detect and populate) and empty folders (provide guided templates).

Extract complete text content from web pages using Claude in Chrome, handling JavaScript-rendered content, lazy-loaded pages, virtual scrolling, SPAs, and canvas-rendered content (Unity WebGL, Unreal, custom WebGL/WebGPU apps). Use this skill whenever the user wants to: scrape a web page, extract text from a URL, save web page content to a file, read a web document, or capture content from any dynamically-loaded website. Trigger especially when WebFetch fails (403, requires auth, JS-rendered) and the user has Chrome connected. Also use when the user mentions extracting content from sites like Google Docs, Notion, 飞书, 轻雀文档, Confluence, Unity WebGL builds, or any SPA/JS-heavy/canvas-heavy site.

Invoke OpenAI Codex CLI (GPT-5.4) non-interactively to get a second opinion, critique, or real-time web search from within a Claude Code session. Use when the user says "ask codex", "call codex", "get codex's opinion", "second opinion from codex", "codex review", or wants to consult Codex on code, architecture, or any technical question. CRITICAL ROUTING RULE: When the user wants Codex to perform a web search (current info, latest versions, recent news), ALWAYS use this skill (--search flag) — never codex:rescue, which cannot do web search. Input: natural-language question, optionally with code context, via stdin heredoc. Output: plain-text analysis from a temp file (or structured JSON via --output-schema). Requires codex-cli installed (`codex` binary in PATH, auth via `codex login`).

Submit a bug report or feature request to the Sentinel-Coding GitHub repository. Use when the user says "submit issue", "report bug", "file issue", "report issue", "feature request", "found a bug", or wants to report a problem with Sentinel-Coding. Requires gh CLI installed and authenticated.

Install Anthropic-internal coding standard hooks into the current project. Copies hook scripts to .claude/hooks/ and merges configuration into .claude/settings.json. Use when the user says "install ant hooks", "add ant mode", "enable coding standards", "install coding hooks", "enhance hooks", or "set up ant prompt".

Remove Anthropic coding standard hooks from the current project. Cleans hook entries from .claude/settings.json and optionally deletes copied scripts. Use when the user says "remove ant hooks", "uninstall ant mode", "disable coding hooks", "remove ant prompt", or "uninstall hooks".

Produce submission-ready documents with compliance lint and format rendering. Use when "export", "sentinel-export", "compliance check", "clean for submission", or when the user wants to render project docs to md/pdf/pptx/docx with AI writing pattern detection.

Generate a structured progress.yaml session entry with typed Candidates for the compaction promotion pipeline, and optionally a durable session report under docs/sessions/. Use at the end of a session, after completing a task, or when the user says "progress", "log session", "what did we learn", "save lessons". Produces entries that compaction can process into CLAUDE.md (rules) and ARCHITECTURE.md (facts).

Scan global skills against project context and produce a tool routing report for developer review. Use when initializing a project, onboarding new tools, or when the developer says "routing", "scan tools", "which skills do I need", "update tool inventory". Produces docs/tool-routing-report.md and STOPS for human approval before any skills are moved.

Propose a Ralph Loop iteration based on current project state. Reads progress.yaml and PRD.md to identify the next task, builds a structured prompt with completion condition, and saves a copy-pasteable /ralph-loop:ralph-loop command. Use when "sentinel-loop", "iterate", "start working", "what's next", or after /boundary completes.

Bootstrap a new project with PRD, ARCHITECTURE, and CLAUDE.md through guided interview. Use when the user says "start", "init project", "set up project", "bootstrap", "new project", or when no PRD.md/ARCHITECTURE.md exists. Produces PRD.md → ARCHITECTURE.md → CLAUDE.md → progress.yaml, then hands off to /routing.