Agent Workflow Skills

Meta's 86M prompt injection and jailbreak detector. Filters malicious prompts and third-party data for LLM apps. 99%+ TPR, <1% FPR. Fast (<2ms GPU). Multilingual (8 languages). Deploy with HuggingFace or batch processing for RAG security.

Autonomous AI agent platform for building and deploying continuous agents. Use when creating visual workflow agents, deploying persistent autonomous agents, or building complex multi-step AI automation systems.

Multi-agent orchestration framework for autonomous AI collaboration. Use when building teams of specialized agents working together on complex tasks, when you need role-based agent collaboration with memory, or for production workflows requiring sequential/hierarchical execution. Built without LangChain dependencies for lean, fast execution.

Build complex AI systems with declarative programming, optimize prompts automatically, create modular RAG systems and agents with DSPy - Stanford NLP's framework for systematic LM programming

> This skill should be used when the user wants to analyze hermesvm binary size changes across a range of commits. Use when the user mentions "binary size", "size analysis", "size regression", "size increase", or asks to measure how commits affect the hermesvm library size.

Generates optimized prompts for AI tools. Activates only when the user explicitly asks to write, fix, improve, or adapt a prompt for a specific AI tool (LLM, Cursor, Midjourney, image AI, video AI, coding agents, etc.). Does not activate for general conversation, coding tasks, document writing, or other non-prompt-engineering work.

Author a new Agent Skill for a Hive agent that conforms to the Agent Skills specification (SKILL.md with YAML frontmatter, optional scripts/references/assets directories). Use when the user asks to create, scaffold, add, or package a new skill for a Hive agent.

Use when you need state across calls — building env vars, navigating with cd, driving REPLs (python -i, mysql, psql, node), or responding to interactive prompts (sudo password, ssh host-key confirmation, mysql connection). Teaches the prompt-sentinel exec pattern (default mode), raw I/O for REPLs (raw_send=True then read_only=True), the one-in-flight-per-session rule, and the close-or-leak-against-the-cap discipline. Bash on macOS — never zsh; explicit shell=/bin/zsh is rejected. Read before calling terminal_pty_open.

Read when a terminal-tools call returned something surprising — empty stdout despite no error, exit_code is null, output_handle came back expired, "too many jobs" / "session busy" / "too many PTYs", warning was set unexpectedly, semantic_status disagrees with exit_code. Diagnostic recipes only — load on demand. Don't preload; the foundational skill covers the happy path.

Build, modify, debug, or review TypeScript MCP servers with mcp-use.

Nightly refinement of an existing per-repo review-style prompt using this reviewer's own finding outcomes. Read confirmed (resolved-by-commit / thumbs-up) and dismissed (thumbs-down) findings, promote the bug patterns the team actually fixes, demote the false-positive patterns, reconcile against the current prompt, and save the refined version. Use this once outcomes exist; use bootstrap-repo-analysis for a cold-start repo.

Read or configure the vault filing methodology and suggest destinations for planned knowledge creation under Generic, LYT, PARA, or Zettelkasten. Use for wiki mode, methodology mode, what is my vault mode, set vault mode, switch to PARA, use LYT, Zettelkasten setup, change mode, configure mode, or methodology routing. This skill does not save content or migrate notes.

Scan an ADK recipe directory and generate a manifest.yaml for it based on the schema at .github/schemas/manifest-schema.json. Use when the user wants to create or generate a manifest.yaml for a recipe under core/, contrib/, or skills/.

Install and authenticate, on demand, the CLIs the sandbox does not prebake — Node/npm, `gws` (Google Workspace), `gcloud`, `agents-cli` (call remote A2A/ADK agents), and `mcp-cli` (use MCP-server tools). Use this whenever one of those tools is needed but missing (a `node`/`npm`/`gws`/`gcloud`/`agents-cli`/`mcp-cli` command returns "command not found"), or before starting any task that requires one — Google Workspace work (Drive, Gmail, Sheets, Calendar, Chat), GCP via `gcloud`, calling another agent deployed remotely over HTTP (Cloud Run or Vertex Agent Runtime), or using tools exposed by an MCP server. Setup only (install + config + headless auth); each tool's own usage lives in its own skill(s).

Helps users discover and install NEW agent skills from external sources (skills.sh, the Skills CLI). Use only when the user wants capabilities they don't already have — "find a skill for X", "is there a skill that can...", "how do I do X" for a task not covered by an installed skill. Do NOT use this to list skills you ALREADY have — those are in your `<available_skills>` block; answer from there directly.

Investigate unexpected chat agent behavior by analyzing direct debug logs in JSONL files. Use when users ask why something happened, why a request was slow, why tools or subagents were used or skipped, or why instructions/skills/agents did not load.

**WORKFLOW SKILL** — Create, update, review, fix, or debug VS Code agent customization files (.instructions.md, .prompt.md, .agent.md, SKILL.md, copilot-instructions.md, AGENTS.md). USE FOR: saving coding preferences; troubleshooting why instructions/skills/agents are ignored or not invoked; configuring applyTo patterns; defining tool restrictions; creating custom agent modes or specialized workflows; packaging domain knowledge; fixing YAML frontmatter syntax. DO NOT USE FOR: general coding questions (use default agent); runtime debugging or error diagnosis; MCP server configuration (use MCP docs directly); VS Code extension development. INVOKES: file system tools (read/write customization files), ask-questions tool (interview user for requirements), subagents for codebase exploration. FOR SINGLE OPERATIONS: For quick YAML frontmatter fixes or creating a single file from a known pattern, edit the file directly — no skill needed.

> Classify a proposed AI use case against your registry — approved, conditional, or not approved — and produce required conditions and next steps. Flags cross-plugin handoffs to privacy or product counsel. Use when user says "triage this use case", "can we use AI for X", "is this approved", "what do we need to do to use AI for X".

> Show contracts with cancel-by deadlines coming up and warn before notice windows close, working from a maintained renewal register. Use when the user asks "what's renewing soon", "what renewals are due", "did we miss a cancellation window", "add this to the renewal tracker", or on a scheduled basis. Receives handoffs from saas-msa-review.

> Review a vendor agreement, NDA, or SaaS subscription against your playbook. Identifies the agreement structure from titles, routes to the right review skill (vendor-agreement-review, nda-review, saas-msa-review), and integrates the output into a single memo. Use when the user says "review this contract", "check this MSA", "is this NDA okay", "look at this SaaS agreement", or attaches an inbound agreement for review.

> Detects when Luminance, Kira, or a similar bulk-review tool is in use, hands off the high-volume clause extraction to it, and QAs its output per the trust level in `~/.claude/plugins/config/claude-for-legal/corporate-legal/CLAUDE.md`. Use when user says "send to Luminance", "bulk review", "AI extraction", or when diligence-issue-extraction hits a high-volume category.

> Update the status of an in-progress international expansion project — recalculates what is now unblocked, flags anything overdue, and surfaces the next priorities. Use when work has happened since the last session and the expansion tracker needs to reflect the current state.

> Grade an IRAC essay for structure, issue-spotting, rule accuracy, analysis depth, and organization. Does NOT rewrite the essay or show a model answer; tracks patterns across sessions. Use when the user says "grade my IRAC", "check my essay", or "I wrote this, give me feedback".

> Check installed community skills for updates. Shows a diff and requires explicit approval before applying. Use when the user says "check for updates", "update my skills", "anything new for my installed skills", or when invoked from the registry-sync agent.

> Run a focused N-question study session on a subject — MBE, essay, or flashcards. Tracks performance and updates the study plan. Use when the user says "run me 10 questions on [subject]", "do a session on [subject]", "let's do 5 cards on [subject]", or wants to drill a fixed number of questions and have the plan adapt.

> Install a community skill from a watched registry. Reads the allowlist first, fetches, shows the RAW SKILL.md (not just a summary), runs structural trust checks, runs skills-qa, and only writes files after explicit user approval. Use when the user says "install [skill]", picks install from browse, or provides a direct skill URL.

> Suggest community skills based on recent activity in other plugins. Checks whether the community has built something relevant to a task and mentions it once, non-intrusively. Use when the user says "is there a community skill for this", "what else is out there", or asks for skill recommendations; also runs passively as part of other plugins' workflows.

> skills installed via the legal builder hub. Safe by default — refuses to touch first-party plugin skills, confirms before removing files, and logs

> End-of-semester case handoff memos — the mirror of /ramp. Produces per-case transition memos and a cohort summary so the departing cohort hands work to the incoming cohort cleanly. Reads deadlines, client-comms, and case history. Use when the professor or departing students need to wrap up the semester, build transition memos, or offboard a graduating/withdrawing student.

Bootstrap skill for discovering additional skills and context from a parent project when workerd is used as a submodule. Load this skill when tasks span project boundaries (e.g., Sentry/production investigation, integration testing, cross-repo debugging).

Detect and fix drift across agent-first infrastructure files. Ensures skill inventories, workflow chains, architecture tables, issue/PR templates, and cross-references stay consistent when skills, crates, or workflows change. Run after adding, removing, or renaming skills or components. Trigger keywords - sync agent infra, sync skills, update agent docs, check agent consistency, agent infra drift, sync contributing, sync agents.

>- Upload local assets (images, mockups, extracted HTML, design markdown) to a Stitch project. ALWAYS use this skill when you need to upload visual assets, HTML pages, or design docs to Stitch, particularly when direct MCP tool calls fail or truncate due to base64 token limits.

Teaches agents to iteratively build websites using Stitch with an autonomous baton-passing loop pattern

当面对 2 个以上可以独立进行、无共享状态或顺序依赖的任务时使用

当在当前会话中执行包含独立任务的实现计划时使用

在开始任何对话时使用——确立如何查找和使用技能，要求在任何响应（包括澄清性问题）之前调用 Skill 工具

当创建新技能、编辑现有技能或在部署前验证技能是否有效时使用

Interact with iOS simulator or Android emulator/device using snapshot-based coordinates. Uses accessibility tree snapshots for precise element targeting, with screenshot verification as fallback. Use when navigating the app on a simulator/emulator.

Analyze agent feedback artifacts from GitHub Actions workflow runs, extract actionable learnings, and incorporate them into skill files and CLAUDE.md. Tracks scan progress to avoid re-processing.

Generate candidate fixes for verified security findings. Consumes TRIAGE.json (preferred), VULN-FINDINGS.json, INCIDENTS.json, or a vuln-pipeline results directory. Pipeline input is delegated to the execution-verified `vuln-pipeline patch` ladder; static-analysis input gets a per-finding patch subagent + independent reviewer and is written as inert diffs for human review. Writes PATCHES/bug_NN/{patch.diff,patch_result.json}, PATCHES.md, and PATCHES.json. Use when asked to "fix the findings", "patch these vulns", "generate fixes", or "close the loop on triage".

>- Static source-code vulnerability scan. Reads a target directory (and THREAT_MODEL.md if present), spawns parallel review subagents per focus area, and writes VULN-FINDINGS.json + .md for /triage to consume. Read-only — no building, running, or network. For execution-verified crashes, use vuln-pipeline instead. Use when asked to "scan for vulns", "review this code for security issues", "find bugs in <dir>", or as the step between /threat-model and /triage.

Orchestrates a strict TDD loop across subagents to build features.

Audits GitHub Actions workflows for security vulnerabilities in AI agent integrations including Claude Code Action, Gemini CLI, OpenAI Codex, and GitHub AI Inference. Detects attack vectors where attacker-controlled input reaches AI agents running in CI/CD pipelines, including env var intermediary patterns, direct expression injection, dangerous sandbox configurations, and wildcard user allowlists. Use when reviewing workflow files that invoke AI coding agents, auditing CI/CD pipeline security for prompt injection risks, or evaluating agentic action configurations.

Iteratively reviews and fixes Claude Code skill quality issues until they meet standards. Runs automated fix-review cycles using the skill-reviewer agent. Use to fix skill quality issues, improve skill descriptions, run automated skill review loops, or iteratively refine a skill. Triggers on 'fix my skill', 'improve skill quality', 'skill improvement loop'. NOT for one-time reviews—use /skill-reviewer directly.

Selects bounded, graph-informed source slices with Trailmark and delegates focused code analysis or patch-proposal work to a smaller subagent. Use when offloading function-, class-, caller-, callee-, call-path-, entrypoint-, or line-focused code tasks to constrained or locally hosted models without exposing the full repository.

Design multi-agent AI workflows with clear boundaries, handoffs, and monitoring. Use when a complex PM task should run as parallel specialized agents instead of one linear process.

Diagnose context stuffing vs. context engineering. Use when an AI workflow feels bloated, brittle, or hard to steer reliably.

Triage a competitive or market question into the right intelligence disciplines, cadence, and executing skill. Use when you know something needs researching but not which channel to run.