Skip to content

SKILL-INDEX.md

Latest commit

 

History

History
921 lines (705 loc) · 40.2 KB

SKILL-INDEX.md

File metadata and controls

921 lines (705 loc) · 40.2 KB
Raw

Skill Index

Use this file to find the right skill for your task. Read the skill's SKILL.md or README before using it.

Routing Protocol

Agents should trigger skills by task intent, not only when the user names a skill.

  1. Classify the work: debugging, README, research audit, security, frontend, browser, MCP, documents, game, writing, memory, or architecture.
  2. Search this index for matching What / When lines.
  3. Read the selected skill's SKILL.md.
  4. Follow the skill's workflow and use bundled scripts/references when relevant.
  5. If no existing skill fits a complex task, search curated resources or public upstreams before inventing a workflow.

Public-safety note: this index is documentation, not a credential store. Do not add API keys, tokens, private paths with secrets, or account state.

Obra-Superpowers

Core development workflows for systematic, high-quality work.

systematic-debugging

  • What: Root cause investigation before proposing fixes; prevents symptom-based patching
  • When: Any bug, test failure, unexpected behavior, or performance problem
  • Path: D:\agent-resources\skills\obra-superpowers\systematic-debugging

test-driven-development

  • What: Write tests first, then implementation; ensures correctness and prevents regressions
  • When: Implementing any feature or bugfix before writing implementation code
  • Path: D:\agent-resources\skills\obra-superpowers\test-driven-development

dispatching-parallel-agents

  • What: Coordinate multiple independent tasks across agents without shared state
  • When: Facing 2+ independent tasks that can be worked on without sequential dependencies
  • Path: D:\agent-resources\skills\obra-superpowers\dispatching-parallel-agents

receiving-code-review

  • What: Evaluate and implement code review feedback with technical rigor
  • When: Receiving code review feedback, especially if unclear or technically questionable
  • Path: D:\agent-resources\skills\obra-superpowers\receiving-code-review

requesting-code-review

  • What: Prepare work for peer review to verify requirements and quality
  • When: Completing tasks, implementing major features, or before merging
  • Path: D:\agent-resources\skills\obra-superpowers\requesting-code-review

using-git-worktrees

  • What: Create isolated workspaces for feature work using native git tools
  • When: Starting feature work that needs isolation from current workspace
  • Path: D:\agent-resources\skills\obra-superpowers\using-git-worktrees

verification-before-completion

  • What: Run verification commands and confirm output before claiming success
  • When: About to claim work is complete, fixed, or passing before committing/PRs
  • Path: D:\agent-resources\skills\obra-superpowers\verification-before-completion

subagent-driven-development

  • What: Delegate work to specialized sub-agents with clear specs and verification
  • When: Complex tasks requiring multiple specialized agents or parallel work
  • Path: D:\agent-resources\skills\obra-superpowers\subagent-driven-development

writing-plans

  • What: Create detailed execution plans before starting implementation
  • When: Starting complex multi-step tasks or unfamiliar work
  • Path: D:\agent-resources\skills\obra-superpowers\writing-plans

executing-plans

  • What: Follow execution plans systematically with progress tracking
  • When: Working through complex multi-step tasks
  • Path: D:\agent-resources\skills\obra-superpowers\executing-plans

Anthropics

Anthropic-specific tools and integrations.

claude-api

  • What: Build, debug, and optimize Claude API / Anthropic SDK apps with prompt caching
  • When: Code imports anthropic/@anthropic-ai/sdk; adding Claude features (caching, thinking, tool use, batch, files)
  • Path: D:\agent-resources\skills\anthropics\claude-api

mcp-builder

  • What: Create high-quality MCP servers for LLM integration with external services
  • When: Building MCP servers in Python (FastMCP) or Node/TypeScript (MCP SDK)
  • Path: D:\agent-resources\skills\anthropics\mcp-builder

webapp-testing

  • What: Test local web applications using Playwright; verify UI, debug behavior, capture screenshots
  • When: Testing frontend functionality or debugging UI behavior
  • Path: D:\agent-resources\skills\anthropics\webapp-testing

web-artifacts-builder

  • What: Create complex claude.ai HTML artifacts with React, Tailwind CSS, shadcn/ui
  • When: Building elaborate multi-component artifacts requiring state management or routing
  • Path: D:\agent-resources\skills\anthropics\web-artifacts-builder

skill-creator

  • What: Create, modify, and optimize skills; run evals and benchmark performance
  • When: Creating new skills, editing existing ones, or optimizing skill descriptions
  • Path: D:\agent-resources\skills\anthropics\skill-creator

doc-coauthoring

  • What: Collaborate on document creation and editing
  • When: Co-authoring documents with Claude
  • Path: D:\agent-resources\skills\anthropics\doc-coauthoring

pdf

  • What: Extract, analyze, and manipulate PDF documents
  • When: Working with PDF files programmatically
  • Path: D:\agent-resources\skills\anthropics\pdf

docx

  • What: Create and manipulate Word documents
  • When: Generating or editing .docx files
  • Path: D:\agent-resources\skills\anthropics\docx

xlsx

  • What: Create and manipulate Excel spreadsheets
  • When: Generating or editing .xlsx files
  • Path: D:\agent-resources\skills\anthropics\xlsx

pptx

  • What: Create and manipulate PowerPoint presentations
  • When: Generating or editing .pptx files
  • Path: D:\agent-resources\skills\anthropics\pptx

canvas-design

  • What: Design visual layouts and graphics
  • When: Creating visual designs or layouts
  • Path: D:\agent-resources\skills\anthropics\canvas-design

frontend-design

  • What: Design and build frontend interfaces
  • When: Creating frontend designs and implementations
  • Path: D:\agent-resources\skills\anthropics\frontend-design

brand-guidelines

  • What: Apply and maintain brand consistency
  • When: Ensuring brand compliance in designs
  • Path: D:\agent-resources\skills\anthropics\brand-guidelines

theme-factory

  • What: Generate and manage design themes
  • When: Creating or customizing design themes
  • Path: D:\agent-resources\skills\anthropics\theme-factory

slack-gif-creator

  • What: Create animated GIFs for Slack
  • When: Generating GIF content for Slack messages
  • Path: D:\agent-resources\skills\anthropics\slack-gif-creator

algorithmic-art

  • What: Generate algorithmic art and visualizations
  • When: Creating generative art or algorithmic visualizations
  • Path: D:\agent-resources\skills\anthropics\algorithmic-art

shadcn/ui (official)

Official shadcn/ui skill + full source mirror. shadcn is a code-distribution platform: components are added as editable source into the project via the shadcn@latest CLI (not a runtime dependency). MIT licensed. Full upstream repo mirrored at D:\agent-resources\repos\shadcn-ui (apps, packages, templates, registry, evals).

shadcn

  • What: Add / search / compose / fix / theme shadcn/ui components and presets; injects live project context (components.json, installed components, aliases, tailwind version) and enforces composition + styling rules (semantic tokens, Field/FieldGroup forms, data-icon, gap over space-y).
  • When: Any project with a components.json; building or upgrading React/Next/Vite UIs; "shadcn init", "add ", "apply --preset", theming, or component debugging. Auto-triggers by task intent, not only when named.
  • Use: run all CLI via the project's package runner (npx/pnpm dlx/bunx --bun shadcn@latest). Read SKILL.md rules/*.md (styling, forms, composition, icons, base-vs-radix) before writing component markup.
  • Path: D:\agent-resources\skills\shadcn (→ repos\shadcn-ui\skills\shadcn)
  • Repo: D:\agent-resources\repos\shadcn-ui

Context-Engineering-Kit

Systematic approaches to context, documentation, and development practices.

tdd

  • What: Test-driven development methodology
  • When: Implementing features with test-first approach
  • Path: D:\agent-resources\skills\context-engineering-kit\tdd

ddd

  • What: Domain-driven design principles
  • When: Designing systems around business domains
  • Path: D:\agent-resources\skills\context-engineering-kit\ddd

sdd

  • What: Specification-driven development
  • When: Implementing from detailed specifications
  • Path: D:\agent-resources\skills\context-engineering-kit\sdd

sadd

  • What: Specification-and-architecture-driven development
  • When: Complex systems requiring both spec and architecture planning
  • Path: D:\agent-resources\skills\context-engineering-kit\sadd

review

  • What: Code review practices and workflows
  • When: Reviewing code or preparing for review
  • Path: D:\agent-resources\skills\context-engineering-kit\review

reflexion

  • What: Reflection and learning from past work
  • When: Analyzing past decisions and outcomes
  • Path: D:\agent-resources\skills\context-engineering-kit\reflexion

kaizen

  • What: Continuous improvement methodology
  • When: Incrementally improving processes or code
  • Path: D:\agent-resources\skills\context-engineering-kit\kaizen

git

  • What: Git workflows and best practices
  • When: Working with version control
  • Path: D:\agent-resources\skills\context-engineering-kit\git

mcp

  • What: Model Context Protocol practices
  • When: Working with MCP servers and tools
  • Path: D:\agent-resources\skills\context-engineering-kit\mcp

docs

  • What: Documentation best practices
  • When: Writing or maintaining documentation
  • Path: D:\agent-resources\skills\context-engineering-kit\docs

fpf

  • What: Fail-fast, fail-forward principles
  • When: Rapid iteration and learning
  • Path: D:\agent-resources\skills\context-engineering-kit\fpf

tech-stack

  • What: Technology stack selection and management
  • When: Choosing or evaluating tech stacks
  • Path: D:\agent-resources\skills\context-engineering-kit\tech-stack

customaize-agent

  • What: Customize agent behavior and capabilities
  • When: Configuring agents for specific tasks
  • Path: D:\agent-resources\skills\context-engineering-kit\customaize-agent

Vipin

Specialized skills for workstation maintenance, Lark, communication, and content creation.

workstation-maintenance

  • What: Safely inventory, classify, organize, and maintain Vipin's Windows workstation across C:/, D:/, and G:/, including file-level approved moves, D-drive root move-with-junction organization, devtools, AGENT_RESOURCE/agent-resources, AGENTIC_SCIENCE/UUPF, WEIPING_WIKI, agent skill sync, preflight, broad-approval execution, and rollback
  • When: Cleaning up drives, organizing local files or D-root directories, protecting D:/Research plus AGENT_RESOURCE, AGENTIC_SCIENCE, devtools, and public devtools exports, preparing type-grouped 30-day-age-gated batch-confirmed file moves, preflighting all batches without moving files, moving eligible D-root directories into D:/_Organized while preserving old paths as junctions, executing broadly approved low-risk batches, or coordinating agent infrastructure upgrades with weiping-wiki public-safe docs
  • Path: D:\agent-resources\skills\vipin\workstation-maintenance

project-hardness

  • What: Turn any project into an AI-readable, maintainable, verifiable causal layer under <project>/.agent/ (architecture, glossary, constraints, modules, flows, entities, causal-graph, decisions, tasks), and harden vague requirements into concrete task specs (goal, modules, edit paths, risks, acceptance criteria, test commands). Engine lives in WEIPING_WIKI (python scripts/wiki.py hardness); facts stay project-local, only cross-project lessons sync to agentmemory
  • When: Starting work on an unfamiliar/large project, needing a change's blast radius or do-not-modify boundaries, or hardening a vague requirement before editing. Applies to every agent
  • Path: D:\agent-resources\skills\vipin\project-hardness

literature-to-manga

  • What: Convert prose, novel, biography, history, or book-archive sources into manga adaptation packages with candidate scoring, source extraction, manga bibles, chapter/page/panel scripts, storyboard briefs, and image prompt packs
  • When: Selecting works from D:\Company\anime\literature-books, adapting local book files into comics or manga, ranking adaptation candidates, or generating character/scene/panel prompt packages from prose
  • Path: D:\agent-resources\skills\vipin\literature-to-manga

lark-cli

  • What: Command-line interface for Lark platform operations
  • When: Automating Lark operations via CLI
  • Path: D:\agent-resources\skills\vipin\lark-cli

lark-base

  • What: Lark Base (database) operations and automation
  • When: Working with Lark Base databases
  • Path: D:\agent-resources\skills\vipin\lark-base

lark-doc

  • What: Lark Doc (document) creation and management
  • When: Creating or managing Lark documents
  • Path: D:\agent-resources\skills\vipin\lark-doc

lark-sheets

  • What: Lark Sheets (spreadsheet) operations
  • When: Working with Lark spreadsheets
  • Path: D:\agent-resources\skills\vipin\lark-sheets

lark-drive

  • What: Lark Drive (file storage) operations
  • When: Managing files in Lark Drive
  • Path: D:\agent-resources\skills\vipin\lark-drive

lark-calendar

  • What: Lark Calendar operations and scheduling
  • When: Managing calendar events in Lark
  • Path: D:\agent-resources\skills\vipin\lark-calendar

lark-contact

  • What: Lark Contact management
  • When: Managing contacts in Lark
  • Path: D:\agent-resources\skills\vipin\lark-contact

lark-im

  • What: Lark Instant Messaging operations
  • When: Sending messages via Lark
  • Path: D:\agent-resources\skills\vipin\lark-im

lark-wiki

  • What: Lark Wiki (knowledge base) operations
  • When: Creating or managing Lark Wiki content
  • Path: D:\agent-resources\skills\vipin\lark-wiki

lark-slides

  • What: Lark Slides (presentation) creation
  • When: Creating presentations in Lark
  • Path: D:\agent-resources\skills\vipin\lark-slides

lark-markdown

  • What: Lark Markdown formatting and conversion
  • When: Working with Markdown in Lark
  • Path: D:\agent-resources\skills\vipin\lark-markdown

lark-openapi-explorer

  • What: Explore and test Lark OpenAPI
  • When: Testing Lark API endpoints
  • Path: D:\agent-resources\skills\vipin\lark-openapi-explorer

lark-shared

  • What: Shared Lark utilities and helpers
  • When: Using common Lark functionality
  • Path: D:\agent-resources\skills\vipin\lark-shared

lark-skill-maker

  • What: Create custom Lark skills
  • When: Building new Lark integrations
  • Path: D:\agent-resources\skills\vipin\lark-skill-maker

feishu-bridge

  • What: Bridge between Feishu and other systems
  • When: Integrating Feishu with external services
  • Path: D:\agent-resources\skills\vipin\feishu-bridge

communication-assistant

  • What: Assist with communication tasks
  • When: Drafting or managing communications
  • Path: D:\agent-resources\skills\vipin\communication-assistant

email-assistant

  • What: Email management and automation
  • When: Handling email tasks
  • Path: D:\agent-resources\skills\vipin\email-assistant

content-creation-publisher

  • What: Create and publish content
  • When: Publishing content to multiple platforms
  • Path: D:\agent-resources\skills\vipin\content-creation-publisher

chrome-automation

  • What: Automate Chrome browser tasks
  • When: Automating browser interactions
  • Path: D:\agent-resources\skills\vipin\chrome-automation

frontend-design

  • What: Frontend design and implementation
  • When: Creating frontend designs
  • Path: D:\agent-resources\skills\vipin\frontend-design

readme-blueprint-generator

  • What: Generate README files from templates
  • When: Creating project documentation
  • Path: D:\agent-resources\skills\vipin\readme-blueprint-generator

wechat-video-channel-publish

  • What: Publish videos to WeChat channels
  • When: Publishing video content to WeChat
  • Path: D:\agent-resources\skills\vipin\wechat-video-channel-publish

paper-orchestra

  • What: Orchestrate academic paper workflows
  • When: Managing academic paper projects
  • Path: D:\agent-resources\skills\vipin\paper-orchestra

Trail of Bits

Security and testing expertise.

agentic-actions-auditor

  • What: Audit agent actions for security and correctness
  • When: Reviewing agent behavior and decisions
  • Path: D:\agent-resources\skills\trailofbits\agentic-actions-auditor

ask-questions-if-underspecified

  • What: Identify and clarify underspecified requirements
  • When: Requirements are vague or incomplete
  • Path: D:\agent-resources\skills\trailofbits\ask-questions-if-underspecified

audit-context-building

  • What: Audit how context is built and used
  • When: Reviewing context management
  • Path: D:\agent-resources\skills\trailofbits\audit-context-building

building-secure-contracts

  • What: Design and implement secure smart contracts
  • When: Building blockchain contracts
  • Path: D:\agent-resources\skills\trailofbits\building-secure-contracts

static-analysis

  • What: Perform static code analysis
  • When: Analyzing code for issues without execution
  • Path: D:\agent-resources\skills\trailofbits\static-analysis

mutation-testing

  • What: Test code quality via mutation testing
  • When: Evaluating test suite effectiveness
  • Path: D:\agent-resources\skills\trailofbits\mutation-testing

property-based-testing

  • What: Generate test cases via property-based testing
  • When: Testing with generated inputs
  • Path: D:\agent-resources\skills\trailofbits\property-based-testing

semgrep-rule-creator

  • What: Create Semgrep rules for code analysis
  • When: Building custom static analysis rules
  • Path: D:\agent-resources\skills\trailofbits\semgrep-rule-creator

variant-analysis

  • What: Analyze code variants for consistency
  • When: Comparing similar code implementations
  • Path: D:\agent-resources\skills\trailofbits\variant-analysis

supply-chain-risk-auditor

  • What: Audit supply chain security risks
  • When: Evaluating dependency and integration risks
  • Path: D:\agent-resources\skills\trailofbits\supply-chain-risk-auditor

constant-time-analysis

  • What: Analyze code for timing vulnerabilities
  • When: Reviewing cryptographic or security-sensitive code
  • Path: D:\agent-resources\skills\trailofbits\constant-time-analysis

differential-review

  • What: Review code changes systematically
  • When: Reviewing pull requests or diffs
  • Path: D:\agent-resources\skills\trailofbits\differential-review

spec-to-code-compliance

  • What: Verify code matches specifications
  • When: Ensuring implementation matches spec
  • Path: D:\agent-resources\skills\trailofbits\spec-to-code-compliance

Standalone

Specialized single-purpose skills.

gsap-core

  • What: GSAP animation library core concepts
  • When: Working with GSAP animations
  • Path: D:\agent-resources\skills\standalone\gsap-core

gsap-react

  • What: GSAP integration with React
  • When: Using GSAP in React projects
  • Path: D:\agent-resources\skills\standalone\gsap-react

gsap-scrolltrigger

  • What: GSAP ScrollTrigger plugin for scroll animations
  • When: Creating scroll-based animations
  • Path: D:\agent-resources\skills\standalone\gsap-scrolltrigger

gsap-timeline

  • What: GSAP Timeline for complex animations
  • When: Orchestrating multiple animations
  • Path: D:\agent-resources\skills\standalone\gsap-timeline

gsap-plugins

  • What: GSAP plugin ecosystem
  • When: Using advanced GSAP plugins
  • Path: D:\agent-resources\skills\standalone\gsap-plugins

gsap-performance

  • What: Optimize GSAP animation performance
  • When: Improving animation performance
  • Path: D:\agent-resources\skills\standalone\gsap-performance

gsap-frameworks

  • What: GSAP integration with frameworks
  • When: Using GSAP with Vue, Angular, etc.
  • Path: D:\agent-resources\skills\standalone\gsap-frameworks

gsap-utils

  • What: GSAP utility functions and helpers
  • When: Using GSAP utilities
  • Path: D:\agent-resources\skills\standalone\gsap-utils

darwin-skill

  • What: Darwin evolutionary algorithm implementation
  • When: Implementing genetic algorithms
  • Path: D:\agent-resources\skills\standalone\darwin-skill

nuwa-skill

  • What: MIT-licensed person/mentor skill distillation workflow ("女娲造人"): deep research -> thinking-frame extraction -> runnable perspective skill.
  • When: The user wants to create, update, or choose a person/topic perspective skill from an explicit name or a vague thinking/decision need.
  • Path: D:\agent-resources\skills\standalone\nuwa-skill

huashu-design

  • What: HTML-native design skill (花叔Design) — high-fidelity prototypes, interactive Demos, slide decks, motion/animation, design-variation exploration, plus a design-direction advisor and expert design review; embodies the matching expert (UX / animator / slide designer / prototyper).
  • When: User asks for UI/UX, prototype/mockup, slide deck/presentation, motion graphics/animation, infographic, or "explore a few design directions"; any visual or design-fidelity work in HTML.
  • Path: D:\agent-resources\skills\standalone\huashu-design

colleague-skill

  • What: Colleague agent collaboration
  • When: Working with colleague agents
  • Path: D:\agent-resources\skills\standalone\colleague-skill

anbeime-skill

  • What: Anbeime platform integration
  • When: Working with Anbeime
  • Path: D:\agent-resources\skills\standalone\anbeime-skill

deepseek-mcp

  • What: DeepSeek MCP server integration
  • When: Using DeepSeek models via MCP
  • Path: D:\agent-resources\skills\standalone\deepseek-mcp

mattpocock-skills

  • What: Matt Pocock's TypeScript and web development skills
  • When: Advanced TypeScript or web development patterns
  • Path: D:\agent-resources\skills\standalone\mattpocock-skills

canon-pianist-ai-video-project

  • What: AI video generation for piano performances
  • When: Creating AI-generated piano videos
  • Path: D:\agent-resources\skills\standalone\canon-pianist-ai-video-project

Composio

Composio automation integrations (1000+ services).

Composio provides automation skills for integrating with external services. Key categories include:

  • Communication: Slack, Discord, Telegram, WhatsApp, Email services
  • Project Management: Jira, Asana, Monday.com, Trello, Linear
  • CRM: Salesforce, HubSpot, Pipedrive, Zoho
  • Finance: Stripe, PayPal, QuickBooks, Xero
  • Cloud: AWS, Google Cloud, Azure, DigitalOcean
  • Data: Airtable, Notion, Google Sheets, Zapier
  • AI/ML: OpenAI, Anthropic, Hugging Face, Replicate
  • Web: Shopify, WordPress, Webflow, Wix
  • Analytics: Google Analytics, Mixpanel, Amplitude
  • And 900+ more services

Path: D:\agent-resources\skills\composio{service-name}-automation

Graphify

Graph visualization and analysis.

graphify

  • What: Create and analyze graph visualizations
  • When: Visualizing network or hierarchical data
  • Path: D:\agent-resources\skills\graphify\graphify

MCP Tools (configured in ~/.claude/mcp.json)

These are always-available MCP servers, not file-based skills. Any agent sharing the CC MCP config can use them directly.

media-gen

  • What: Generate images via gpt-image-2 (OpenAI), saves to local disk
  • When: Need to create illustrations, diagrams, banners, paper figures, any image generation task
  • How: MCP tool call (auto-available in CC/ARIS sessions), uses 9100 rotator

agentmemory

  • What: Persistent shared memory — search, recall, lessons, signals, actions
  • When: Always (session start search, cross-agent communication, saving lessons)
  • How: MCP tool calls (memory_smart_search, memory_recall, memory_save, etc.)

codex (reviewer)

  • What: Invoke Codex/GPT-5.5 as an external reviewer
  • When: Research review, code audit, getting a second opinion from a different model
  • How: MCP tool call, used by ARIS /research-review and /auto-review-loop skills

ECC — Agent Harness Performance System

232 skills + 60 agents for Claude Code, Codex, Cursor, OpenCode, Gemini, and more. The most comprehensive open-source agent harness available.

Install (plugin, recommended):

/plugin marketplace add https://github.com/affaan-m/ECC
/plugin install ecc@ecc

Key commands: /ecc:plan, /code-review, /security-scan, /build-fix, /refactor-clean, /test-coverage, /instinct-status, /evolve, /harness-audit

Components at D:\agent-resources\skills\ecc\:

  • skills/ — 232 workflow skills (TDD, security, frontend, backend, ML, agent-harness, etc.)
  • agents/ — 60 specialized subagents (architect, code-reviewer, security-reviewer, tdd-guide, planner, etc.)
  • rules/ — always-on coding guidelines (common + language-specific: TS, Python, Go, Swift, PHP)
  • hooks/ — event-driven automations (tool use, session start/stop, file edits)
  • mcp-configs/ — MCP server definitions (GitHub, Supabase, Vercel, etc.)
  • COMMANDS-QUICK-REF.md — full command reference
  • the-shortform-guide.md — quick-start guide

Notable features:

  • AgentShield: 102-rule security scanner for CLAUDE.md, hooks, MCP configs
  • Continuous Learning v2: instinct extraction with confidence scoring
  • Dashboard GUI: python3 ecc_dashboard.py or npm run dashboard
  • Hook tuning via env vars: ECC_HOOK_PROFILE=minimal|standard|strict

Path: D:\agent-resources\skills\ecc\

Four-piece deployment (四件套, 2026-06-08): ECC is wired into the agent family across four pillars — see D:\devtools\AGENTS.md for the per-agent matrix.

  • Skills: each agent's skills\ecc is now a junction to D:\agent-resources\skills\ecc (CC + Codex converted from drifted copies; old copies archived *.prejxn). Do NOT run ECC's install-apply.js against a live home — it does copyFileSync and re-creates drift; junction by hand instead.
  • Intuition: D:\devtools\agent-hooks\ecc-observe.cmd (node observe-runner.js, NOT observe.sh) wired additively into CC settings.json; default OFF via marker, toggle ecc-intuition-toggle.cmd. CLV2_HOMUNCULUS_DIR=/d/devtools/data/derived/clv2-homunculus, CLAUDE_PLUGIN_ROOT=D:\agent-resources\repos\ECC (full repo, has scripts/).
  • Memory: ECC homunculus is a derivation layer only → data\derived\lessons\pending\ecc\ → promoted by scripts\ingest-lessons.mjs. agentmemory stays the single source.
  • Security: pinned ecc-agentshield@1.4.0 at D:\devtools\agent-tools\agentshield\, run via bin\agentshield-scan.cmd (baseline-delta, nightly), never inlined into hooks.

hermes-agent (family member, not a skill pack)

NousResearch self-improving agent — ambient/background role (gateway, cron, lesson curation). Full independent Python agent, NOT a skill pack.

  • Source: D:\devtools\vendor\hermes-agent; venv + home: D:\devtools\hermes (Python 3.11); launcher: D:\devtools\hermes.cmd (pins HERMES_HOME to D-drive).
  • Model: gpt-5.5 relay (中转) via hermes\config.yaml. agentmemory MCP wired; native memory/session_search demoted. agentskills.io-compatible.
  • See D:\devtools\AGENTS.md for full wiring.

Karpathy Guidelines

Behavioral guidelines derived from Andrej Karpathy's observations on LLM coding pitfalls.

karpathy-guidelines

  • What: Four coding principles — Think Before Coding, Simplicity First, Surgical Changes, Goal-Driven Execution
  • When: Writing, reviewing, or refactoring any code to avoid overcomplication and scope creep
  • Path: D:\agent-resources\skills\standalone\karpathy-guidelines

Understand-Anything

Turn any codebase or knowledge base into an interactive knowledge graph.

understand

  • What: Build a knowledge graph from a codebase using Tree-sitter + LLM semantic analysis
  • When: Starting work on an unfamiliar codebase; need architecture overview
  • Path: D:\agent-resources\skills\standalone\understand

understand-chat

  • What: Ask natural language questions about the codebase knowledge graph
  • When: Exploring how components relate, tracing data flows, answering "where is X"
  • Path: D:\agent-resources\skills\standalone\understand-chat

understand-dashboard

  • What: Open the interactive visual knowledge graph explorer
  • When: Need a visual map of the codebase structure
  • Path: D:\agent-resources\skills\standalone\understand-dashboard

understand-diff

  • What: Analyze the impact of current changes on the codebase graph
  • When: Before committing; want to know what a change touches
  • Path: D:\agent-resources\skills\standalone\understand-diff

understand-explain

  • What: Deep-dive explanation of a specific file or function
  • When: Need to understand a complex module in detail
  • Path: D:\agent-resources\skills\standalone\understand-explain

understand-onboard

  • What: Generate an onboarding guide for new team members
  • When: Onboarding someone to a codebase; writing contributor docs
  • Path: D:\agent-resources\skills\standalone\understand-onboard

understand-domain

  • What: Extract business domain knowledge (domains, flows, steps) from code
  • When: Need to understand business logic embedded in the codebase
  • Path: D:\agent-resources\skills\standalone\understand-domain

understand-knowledge

  • What: Analyze a Karpathy-pattern LLM wiki into a knowledge graph
  • When: Working with a wiki-style knowledge base (markdown files)
  • Path: D:\agent-resources\skills\standalone\understand-knowledge

Presentation as Code

Create high-impact, interactive HTML presentations through AI collaboration.

presentation-as-code

  • What: Build standalone HTML slide decks with shared design system, data viz, offline-first
  • When: Creating training materials, pitch decks, tech talks, or any visual storytelling
  • Path: D:\agent-resources\skills\standalone\presentation-as-code

CodeGraph (MCP Tool)

Pre-indexed code knowledge graph for AI agents — fewer tokens, fewer tool calls, 100% local.

codegraph-add-lang

  • What: Add a new language to CodeGraph's tree-sitter parser
  • When: Working with a language not yet supported by CodeGraph
  • Path: D:\agent-resources\skills\standalone\codegraph-add-lang

codegraph-agent-eval

  • What: Evaluate agent performance against CodeGraph benchmarks
  • When: Benchmarking or comparing agent behavior on codebase queries
  • Path: D:\agent-resources\skills\standalone\codegraph-agent-eval

Claude Plugins Official (Anthropic)

Official Anthropic-managed plugins for Claude Code workflows.

code-review (plugin)

  • What: Automated PR code review using multiple specialized agents with confidence scoring
  • When: Reviewing pull requests; want multi-agent review with structured output
  • Path: D:\agent-resources\skills\claude-plugins-official\code-review

feature-dev (plugin)

  • What: End-to-end feature development workflow with agents and commands
  • When: Implementing a new feature from spec to code
  • Path: D:\agent-resources\skills\claude-plugins-official\feature-dev

code-modernization (plugin)

  • What: Modernize legacy code to current language/framework standards
  • When: Upgrading old codebases, migrating deprecated APIs
  • Path: D:\agent-resources\skills\claude-plugins-official\code-modernization

pr-review-toolkit (plugin)

  • What: Comprehensive PR review toolkit with agents and commands
  • When: Thorough pull request review with structured checklist
  • Path: D:\agent-resources\skills\claude-plugins-official\pr-review-toolkit

agent-sdk-dev (plugin)

  • What: Build agents using the Claude Agent SDK
  • When: Creating custom agents with the Anthropic Agent SDK
  • Path: D:\agent-resources\skills\claude-plugins-official\agent-sdk-dev

security-guidance (plugin)

  • What: Security review and guidance for code changes
  • When: Security-sensitive code; want automated security analysis
  • Path: D:\agent-resources\skills\claude-plugins-official\security-guidance

code-simplifier (plugin)

  • What: Simplify and reduce complexity in existing code
  • When: Code is too complex, over-engineered, or hard to read
  • Path: D:\agent-resources\skills\claude-plugins-official\code-simplifier

commit-commands (plugin)

  • What: Structured git commit workflow with conventional commits
  • When: Creating commits with consistent format and quality
  • Path: D:\agent-resources\skills\claude-plugins-official\commit-commands

claude-code-setup (plugin)

  • What: Set up Claude Code for a new project (CLAUDE.md, settings, hooks)
  • When: Initializing Claude Code in a new repo
  • Path: D:\agent-resources\skills\claude-plugins-official\claude-code-setup

claude-md-management (plugin)

  • What: Manage and maintain CLAUDE.md files across a project
  • When: Keeping CLAUDE.md up to date as the project evolves
  • Path: D:\agent-resources\skills\claude-plugins-official\claude-md-management

mcp-tunnels (plugin)

  • What: Set up MCP tunnels for remote/cloud MCP server access
  • When: Connecting to MCP servers over network tunnels
  • Path: D:\agent-resources\skills\claude-plugins-official\mcp-tunnels

hookify (plugin)

  • What: Create and manage Claude Code hooks with writing-rules skill
  • When: Setting up automated hooks for Claude Code events
  • Path: D:\agent-resources\skills\claude-plugins-official\hookify

session-report (plugin)

  • What: Generate structured session reports summarizing work done
  • When: End of a work session; want a summary of changes and decisions
  • Path: D:\agent-resources\skills\claude-plugins-official\session-report

build-mcp-server (skill)

  • What: Build a production-ready MCP server from scratch
  • When: Creating a new MCP server (more detailed than mcp-builder)
  • Path: D:\agent-resources\skills\standalone\build-mcp-server

build-mcp-app (skill)

  • What: Build an application that consumes MCP servers
  • When: Creating a client app that uses MCP tools
  • Path: D:\agent-resources\skills\standalone\build-mcp-app

session-report (skill)

  • What: Generate a structured end-of-session report
  • When: Wrapping up a work session; want a summary for handoff
  • Path: D:\agent-resources\skills\standalone\session-report

hookify-writing-rules (skill)

  • What: Rules for writing effective Claude Code hooks
  • When: Creating hooks; want best practices for hook scripts
  • Path: D:\agent-resources\skills\standalone\hookify-writing-rules

Academic Research Skills

Comprehensive Claude Code skill suite for the full academic research pipeline — from ideation to publication.

ars-plan

  • What: Socratic dialogue to walk through paper structure and research plan
  • When: Starting a new paper; need to clarify research question, method, and contribution
  • Path: D:\agent-resources\repos\academic-research-skills\academic-pipeline

ars-write

  • What: Human-in-the-loop writing assistance with style calibration and quality checks
  • When: Drafting or revising paper sections; want AI to match your voice, not replace it
  • Path: D:\agent-resources\repos\academic-research-skills\academic-paper

ars-review

  • What: Automated paper review with integrity gates (hallucination, citation, logic checks)
  • When: Pre-submission review; want a structured checklist before submitting
  • Path: D:\agent-resources\repos\academic-research-skills\academic-paper-reviewer

ars-deep-research

  • What: Deep literature search and synthesis with trust-chain citation provenance
  • When: Literature review; need to find, verify, and synthesize related work
  • Path: D:\agent-resources\repos\academic-research-skills\deep-research

AI Engineering from Scratch

435-lesson, 20-phase open curriculum for building AI systems end-to-end (Python, TypeScript, Rust, Julia). Every lesson ships a reusable artifact.

ai-engineering-curriculum

  • What: Structured curriculum from ML fundamentals to production agents and MCP servers
  • When: Learning AI engineering systematically; building reference implementations
  • Path: D:\agent-resources\repos\ai-engineering-from-scratch\phases

ai-engineering-projects

  • What: Hands-on project templates for each curriculum phase
  • When: Need a starter project for a specific AI engineering topic
  • Path: D:\agent-resources\repos\ai-engineering-from-scratch\projects

OpenHuman

Open-source personal AI super-intelligence harness — local memory, managed services, human-in-the-loop design.

openhuman

  • What: Personal AI assistant with local memory, tool integrations, and human-centered UX
  • When: Building or extending a personal AI assistant; need a reference harness architecture
  • Path: D:\agent-resources\repos\openhuman

Game Development

Complete game development pipeline — from brainstorming to release.

Claude-Code-Game-Studios (73 skills, 49 agents)

  • What: Full game studio hierarchy — directors, leads, specialists for Godot/Unity/UE5
  • When: Building any game; need structured workflow (brainstorm → design → architecture → sprint → QA → release)
  • Key skills: /start, /brainstorm, /create-architecture, /sprint-plan, /prototype, /vertical-slice, /playtest-report, /release-checklist
  • Path: D:\agent-resources\skills\game-studios\
  • Public status: local-only pending provenance/license clearance; do not route implicitly from the public index until cleared.

Godogen (autonomous game generation)

  • What: Autonomous game generation pipeline with screenshot-based self-repair
  • When: Want AI to generate a complete game from a description (Godot 4 / Bevy / Babylon.js)
  • Key feature: Frame-grounded self-repair — judges progress from screenshots, not just compilation
  • Path: D:\agent-resources\skills\godogen\
  • Public status: local-only pending provenance/license clearance; do not route implicitly from the public index until cleared.

GSD-2 (spec-driven development, 58.9k stars)

  • What: Meta-prompting system that prevents context rot; breaks work into atomic plans
  • When: Any complex project where "just prompting" produces bad results; need spec → plan → task → implement flow
  • Note: Full runtime (not just skills) — reference for methodology. Use its workflow pattern.
  • Path: D:\agent-resources\repos\gsd-2\

InkOS

Autonomous novel writing AI agent with multi-agent pipeline and web workbench.

inkos

  • What: Autonomous fiction writing — multi-agent pipeline (Radar→Planner→Composer→Architect→Writer→Observer→Reflector→Normalizer→Auditor→Reviser), 33-dimension auditing, style imitation, fan fiction, AIGC detection, EPUB export, InkOS Studio web UI
  • When: Creative fiction writing, novel generation, short fiction packages, cover generation, chapter continuation/import, style analysis
  • Path: D:\agent-resources\skills\inkos

Agency Agents (agency-agents-zh) — 215+ 中文专家子代理

215+ 个即插即用的中文 AI 专家角色（工程/设计/营销/产品/游戏/安全/金融等 18 部门）。 作为 Claude Code 子代理安装，不是 skill。详见 AGENCY-AGENTS.md。

agency-agents

  • What: 211 个领域专家子代理（前端/后端/安全/DBA/SRE/UI/UX/小红书/抖音/跨境电商/政务/医疗合规/Qt 上位机 等），每个带独立人设、流程、交付物
  • When: 需要某个具体领域的专家视角时；用 /<slug> 调用（如 /engineering-security-engineer）或让主 agent 自动路由
  • 分层: T0 源仓库 / T0.5 slug 化无 BOM 暂存 / T1 ~30 精选常驻全局 / T2 按需灌入项目
  • 安装: install-agency（精选常驻）· install-agency -Category <x> · install-agency -All -Project .（按需）
  • 关键: 子代理 name 必须英文 slug，源仓库是中文 name，勿直接复制原始文件，必经暂存 slug 化
  • 源: D:\agent-resources\repos\agency-agents-zh · 暂存: D:\agent-resources\staging\agency-claude
  • 脚本: D:\devtools\agency-build-staging.ps1 · install-agency.ps1 · agency-fix-integration-bom.ps1

Quick Reference

For comprehensive harness: ECC (232 skills + 60 agents at skills/ecc/) For game development: GSD-2 (spec-driven). The local game-studios and godogen packs are useful for inspection but are not public-routed until provenance/license clearance is complete. For debugging: systematic-debugging For features: test-driven-development or feature-dev (plugin) For parallel work: dispatching-parallel-agents For code review: requesting-code-review, receiving-code-review, or code-review (plugin) For Claude API: claude-api For MCP servers: mcp-builder, build-mcp-server, or build-mcp-app For testing web apps: webapp-testing For complex artifacts: web-artifacts-builder For Lark integration: lark-* skills For security review: Trail of Bits skills or security-guidance (plugin) For external services: Composio skills For presentations: presentation-as-code For codebase understanding: understand-* skills or codegraph For coding discipline: karpathy-guidelines For session wrap-up: session-report For hooks: hookify or hookify-writing-rules For academic research: academic-research-skills For AI engineering curriculum: ai-engineering-from-scratch For personal AI assistant: openhuman For a domain-expert persona (211 roles): agency-agents — /<slug> e.g. /engineering-security-engineer; install via install-agency