File-based planning for complex tasks. Use persistent markdown files as working memory to survive context resets. Creates task_plan.md, findings.md, and progress.md. Use for any task requiring >5 tool calls, research projects, or multi-step implementations.
GPU VRAM management patterns for sharing memory across services (Ollama, Whisper, ComfyUI). OOM retry logic, auto-unload on idle, and service signaling protocol.
WHAT: Guide for creating effective AI agent skills - modular packages that extend Claude's capabilities with specialized knowledge, workflows, and tools. WHEN: User wants to create, write, author, or update a skill. User asks about skill structure, SKILL.md format, or how to package domain knowledge for AI agents. KEYWORDS: "create a skill", "make a skill", "new skill", "skill template", "SKILL.md", "agent skill", "write a skill", "skill structure", "package a skill"
> [fkparse](./index.md) > 技能
Apply task-specific templates to AI session plans using ai-update-plan. Use when starting a new task to load appropriate plan structure (feature, bugfix, refactor, documentation, security).
[Architecture] Use when designing or implementing cross-service communication, data synchronization, or service boundary patterns.
Command-line task management tool for AI coding agents and humans. Provides local SQLite-based task boards for tracking work items, checklists, and comments without requiring an external API. Use this skill when: - Managing multi-step coding tasks - Tracking progress on complex implementations - Breaking down work into checkable subtasks - Documenting decisions and progress via comments - Coordinating work across agent sessions
Maintain Conductor-compatible project context, tracks, specs, and plans.
Automate npm Trusted Publisher setup via the npm-trustme CLI. Use when configuring or verifying npm Trusted Publishers for GitHub Actions with npx npm-trustme, including browser automation and WebAuthn passkey approval.
Scaffold a new skill following best practices.
Apply Anthropic's production multi-agent research patterns for complex research tasks. Use when research involves multiple independent dimensions, requires synthesis across many sources, or needs parallelization for speed. Includes query complexity assessment, parallel execution, progressive search refinement, thinking strategies, and findings compression.
Create or edit a Cline Skill with guided assistance and best practices
Implement APCA/WCAG2 contrast requirements and solver behavior per src/planning/spec-v0.3.md and src/planning/roadmap-v0.3.md, with minimal tests.
Implement an opence change and keep tasks in sync.
Build multi-step LLM agents for the Continuum system. Use when creating agent definitions with nodes, edges, tools, and execution flows. Triggers on requests to "create an agent", "build a workflow", "design an agent graph", or when discussing Continuum agent architecture. Supports parallel execution, conditional routing, tool integration, and structured outputs.
Inspect inventory, stockpile, altars, hearts, and respawn logic.
Adjust biome or terrain distribution parameters.
Map include chains and explain module relationships.
Open the main step loop and action handlers.
Make the trading hub feel more organic without blocking roads.