Agent Workflow Skills

Multi-agent research skill for parallel research execution (10 agents, battle-tested with real case studies).

Expert prompt engineer specializing in advanced prompting techniques, LLM optimization, and AI system design. Masters chain-of-thought, constitutional AI, and production prompt strategies. Use when building AI features, improving agent performance, or crafting system prompts.

Three.js skills for creating 3D elements and interactive experiences

Use whenever the user wants to find, rank, or shortlist directories and listing sites where they can submit an AI product — an AI tool, AI app, AI agent, or agent skill / plugin — to get backlinks, referral traffic, and discovery. Triggers on "where can I list my AI tool", "directories to submit my AI agent", "agent-skills directories", "best AI tool directories for backlinks", "where do I get my GPT/Claude app discovered", or "pull submission details for these AI-directory domains", even when described indirectly (we built an AI agent, where do we get it in front of people). Drives the ServiceGraph API (api.servicegraph.co) — a catalog of 1,000+ product directories enriched with Domain Rating, backlinks, and organic traffic. Defer to find-mcp-directories for MCP-server listings specifically, and to find-product-directories for general SaaS/software/app launches with no AI angle. Skip finding an AI consultancy/agency to hire (use find-ai-consultancy), comparing AI products ("ChatGPT vs Claude"), building an AI tool (do-the-work), and AI link-building *services*.

Use whenever the user wants to find, rank, or shortlist directories and registries where they can submit or list an MCP server (Model Context Protocol server) — to get backlinks, referral traffic, and discovery by agent builders. Triggers on "where do I list my MCP server", "best MCP directories", "MCP registries to submit to", "get my MCP server discovered", or "pull submission details for these MCP-directory domains", even when described indirectly (we built an MCP server, where do we publish it). Drives the ServiceGraph API (api.servicegraph.co) — a catalog of 1,000+ product directories enriched with Domain Rating, backlinks, and organic traffic. Defer to find-ai-directories for general AI-tool / AI-agent / agent-skill listings, and to find-product-directories for general SaaS/software launches. Skip building an MCP server or asking how MCP works (DIY), finding a firm to build one (use find-ai-consultancy / find-software-developer), and MCP link-building *services*.

Use whenever the user wants to find, rank, or shortlist directories, listing sites, or launch platforms where they can submit a software product, SaaS, app, tool, or startup — to get backlinks, referral traffic, and launch-day reach. Triggers on "where can I submit my SaaS for launch", "list of Product Hunt alternatives", "directories to get backlinks for our app", "high-authority software listing sites", "where should I list my startup", or "pull submission details for these 8 directory domains", even when described indirectly (we're launching next week, where do we post; how do we get backlinks by listing in directories). Drives the ServiceGraph API (api.servicegraph.co) — a catalog of 1,000+ product directories enriched with Domain Rating, backlinks, and organic traffic. Defer to find-mcp-directories for MCP-server listings and find-ai-directories for AI-tool / AI-agent / agent-skill listings. Skip finding a firm/agency to hire (use the find-* agency skills), finding products *inside* a directory ("recommend the best CRM"), building a directory site (do-the-work), local/business directories for brick-and-mortar listings, and link-building *services*.

Report active Codex and Claude Code sessions in the current repository by default, with optional machine-wide detail.

Retrospect on a task against local Codex/Claude Code transcripts; propose durable fixes (AGENTS.md, skills).

>- Use for current official documentation about Codex, Codex CLI, or Claude Code behavior, configuration, prompting, skills, permissions, tools, surfaces, capabilities, or troubleshooting; fetch the relevant official URL before answering.

Orchestrate read-only Explore research subagents during planning and one to five Haiku, Sonnet, or Opus subagents to implement the approved plan.

Orchestrate read-only Codex research agents during planning and one to five Codex CLI agents to implement the approved plan.

Use to create/scaffold/init a new agent skill under `.agents/skills` in the working directory where invoked.

Use to find, compare, and recommend current tools, packages, libraries, CLIs, VSCode extensions, agent skills, databases, or infrastructure options for a development task.

"Use to find agent skill installs, duplicate skills, cross-dependencies, invocations, and cross-references across the local machine."

"Use to audit Agent Skills catalogs or installed skill roots for metadata and doc-link issues; optionally apply conservative --fix-safe repairs."

Create decision-complete single- or multi-repository task plans under `.ai/task-handoffs/` and return commands for fresh interactive Codex sessions.

Import memories from other AI memory systems into Cortex. Supports claude-mem (SQLite), Claude Desktop sessions, ChatGPT web export (JSON), Gemini Takeout (JSON), Cursor conversations, and Claude Code JSONL. Use when the user says 'import from claude-mem', 'migrate memories', 'import ChatGPT history', 'import from Gemini', 'transfer memories', or when Cortex detects another memory system is installed.

Bootstrap Cortex for a new project or import existing session history. Use when the user says 'set up Cortex', 'seed this project', 'import my history', 'backfill memories', 'bootstrap memory', 'initialize Cortex for this project', or when starting to use Cortex on an existing codebase that already has Claude Code conversation history.

Use this skill when a designer needs to prepare for a business kickoff workshop with a new customer, or needs to fill a Business Model Canvas after the call. Triggers when someone mentions a kickoff call, BMC workshop, business model canvas, or needs to prepare questions for a new customer discovery session.

>- Create Cursor Agent Skills. Use when authoring a new skill or asking about SKILL.md structure.

> Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

Writes, refactors, and evaluates prompts for LLMs — generating optimized prompt templates, structured output schemas, evaluation rubrics, and test suites. Use when designing prompts for new LLM applications, refactoring existing prompts for better accuracy or token efficiency, implementing chain-of-thought or few-shot learning, creating system prompts with personas and guardrails, building JSON/function-calling schemas, or developing prompt evaluation frameworks to measure and improve model performance.

Audits the bundled `clerk-expo` skill against the @clerk/expo SDK source and clerk-docs, then proposes or applies updates. Use when the user says "audit the expo skill", "update the clerk-expo skill", "check clerk-expo against the SDK", "resync clerk-expo skill", "run audit-expo-skill", or after @clerk/expo ships a minor or major release.

Audits the bundled `clerk-cli` skill against the Clerk CLI source tree and proposes or applies updates. Use when the user says "audit the clerk-cli skill", "update the clerk-cli skill", "check clerk-cli against the CLI source", "resync clerk-cli skill", "run audit-clerk-skill", or after Clerk CLI commands, flags, or agent-mode behavior change.

Set up Hierarchical Agent Memory (HAM) — scoped CLAUDE.md files per directory that reduce token spend. Trigger on "go ham", "set up HAM", "ham commands", "ham help", "ham route", "ham remove", "ham update", "ham status", "ham benchmark", "ham baseline start", "ham baseline stop", "ham metrics clear", "HAM savings", "HAM stats", "HAM dashboard", "HAM sandwich", "HAM insights", "HAM carbon", or "ham sync".

Set up and maintain Hierarchical Agent Memory (HAM) for Codex using scoped local memory files. Trigger on "go ham", "set up HAM", "ham commands", "ham help", "ham route", "ham remove", "ham update", "ham status", "ham benchmark", "ham baseline start", "ham baseline stop", "ham metrics clear", "HAM savings", "HAM stats", "HAM dashboard", "HAM sandwich", "HAM insights", "HAM carbon", or "HAM audit".

Route OpenSkills development tasks to the right project skill or subagent, including sequencing rules for debugging, feature work, regression checks, and release readiness.

Create and refine OpenSkills-compatible skills (SKILL.md + optional resources) with strong metadata, clear activation triggers, and reliable execution guidance.

Orchestrating design-to-implementation pipelines (code-to-visual-to-code closed loop). Persists a project design system across downstream agents. Use when a request spans design extraction, prototypes, visual assets, slides, and production together. Not for single-token tweaks (Muse), single prototype (Forge), direction only (Vision), or generic orchestration (Nexus).

Designing new skill agents via gap analysis, overlap detection, SKILL.md + reference generation, and Nexus integration. Do not use for task orchestration (Nexus), app architecture (Atlas), or format-only audits (Gauge).

Auditing skill/plugin/MCP supply chains. Treats SKILL.md, bundled scripts, MCP server defs, hooks, and `.claude/` config as third-party software. Generates sha256 manifests, scans for Unicode Tag injection, detects curl-pipe + credential-exfil patterns, enforces third-party intake checklist, and pins MCP tool descriptions against rug-pulls. Use when auditing skill/MCP supply chain. Not for app SAST (Sentinel), CI/CD (Gear/Pipe), hook design (Latch), SKILL.md format (Gauge), or runtime exploit (Probe).

Navigating the skill ecosystem and guiding onboarding. Lists agents, recommends best fit for tasks. Don't use for task execution (Nexus), agent design (Architect).

Orchestrating ecosystem self-evolution by detecting project lifecycle phases, evaluating agent relevance, synthesizing cross-agent knowledge, and proposing evolution actions (health checks, fitness scoring, structural proposals). Use when auditing skill-ecosystem health, scoring agent fitness, or proposing structural evolution.

Auditing and optimizing AI CLI configuration. Audits Codex CLI (~/.codex/), Antigravity CLI (~/.gemini/ — `agy`), and Claude Code (~/.claude/) configs (config.toml/settings.json/CLAUDE.md/hooks/MCP) and proposes Before/After diff improvements. Never edits configs directly. Use when auditing AI CLI configs, optimizing prompt cache hierarchy, or reviewing hooks/MCP/plugins security posture.

Reviewing code via multi-engine orchestration. Orchestrates parallel review via subagents with grounding verification, shipping only findings worth fixing. Reviews on three axes — secure, correct, and lean. Default Claude + Codex (dual-engine); agy optional third axis. Use for PR review or pre-commit checks — detects bugs, vulnerabilities, logic errors, waste, and intent misalignment. Offers an interactive pair-review mode (fix-as-you-go). Complements Zen refactoring.

Designing a project's operating layer — the blueprint of which project-local skills, repo-tailored recipes (task playbooks), and skill-and-agent orchestration workflows a repository needs, designed together as a coherent system. Analyzes the repo, designs the layer, then delegates skill-body authoring to Sigil and runtime routing to Nexus. Use when a repository needs a coordinated set of project agents, recipes, and orchestration workflows designed together. Don't use for authoring a single skill body (Sigil), runtime request orchestration (Nexus), autonomous loop runners (Orbit), application or business-process workflows and code-level state machines (Weave), or ecosystem-wide global agents (Architect).

Orchestrating specialist AI agent teams as a meta-coordinator. Decomposes requests into minimum viable chains, spawns each as an independent session in AUTORUN modes, and drives to final output. Use when a task spans multiple specialist domains, requires parallel agent execution, or needs hub-and-spoke routing across the skill ecosystem.

Designing LLM-optimized folder structures. Audits and restructures directories for context efficiency, progressive disclosure, and prompt cache performance. Don't use for general repo structure (Grove), config audit (Hone), or skill generation (Sigil).

Designing and evaluating AI/ML systems across prompt engineering, RAG design, LLM application patterns, AI safety, evaluation frameworks, MLOps, and cost optimization. Use when designing AI/ML pipelines, RAG architectures, prompt strategies, evaluation harnesses, or LLM cost models.

Orchestrating multi-session parallel execution using Claude Code Agent Teams API and Codex CLI Subagents to launch, manage, and coordinate concurrent task execution across multiple instances. Use when parallel work is needed.

Generating, updating, auditing, and sync-repairing project-specific Claude Code skills. Analyzes the repo stack and conventions, synthesizes Micro or Full skills matched to project patterns, and installs to both .claude/skills/ and .agents/skills/. Use when authoring project-local skills.

> AGENTS.md, HANDOFF.md, docs/*) into ONE canonical CLAUDE.md, then reconciles it against the current code state so a fresh session or new teammate can pick up immediately. CLAUDE.md keeps a mandatory PROGRESS "/wrap", "tidy docs", "sync docs", "update docs", "archive", "this phase is done", or the Chinese equivalents "收尾", "同步文档", "整理文档", "整理一下", "更新文档", "存档", "梳理一下", "这阶段做完了". A bare "tidy" / "整理" with prior dev context counts. Do not under-trigger.

Load minimum necessary context into agent context windows. Prevents token bloat, reduces cost, and improves focus. Only load what the current task needs.

Patterns for Retrieval-Augmented Generation (RAG) and agent memory systems. Retrieves only relevant context, prevents context bloat, and maintains coherent state across sessions.

Designs and coordinates multi-agent pipelines where specialized agents collaborate to complete complex tasks. Includes communication protocols, failure handling, and state management.

Guards AI agents and LLM-powered applications against prompt injection attacks — both direct and indirect. Validates AI inputs and outputs at every trust boundary.

Generate 3D models, textures, images, rig characters, animate them, and prepare for 3D printing using the Meshy AI API. Handles API key detection, task creation, polling, downloading, and full 3D print pipeline with slicer integration. Use when the user asks to create 3D models, convert text/images to 3D, texture models, rig or animate characters, 3D print a model, or interact with the Meshy API. For Claude Code or Cursor, use the meshy-3d-generation and meshy-3d-printing skills instead.

> Создаёт полноценных persistent skill-агентов с собственной памятью, system prompt-ом и data-файлами. Автоматизирует создание агентов как у OpenClaw (Doctor, AutoMechanic и т.д.).