Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Manage persistent file-based memory by indexing cross-session user preferences, feedback, and architectural constraints in structured MEMORY.md stores. Use to record user feedback, store project conventions, recall cross-session architectural constraints, or index durable project memory—even if the user doesn't explicitly say "fable-memory" (e.g. "remember this preference", "save to memory", "what are my project rules", "recall user feedback"). Do NOT use for volatile task state stored in .fable/state.json.
A step-by-step problem-solving loop (classify the ask, define done, gather evidence, decide, act surgically, verify by observation, report outcome-first). Use when the user says "/fable-method", "use the fable method", or "approach this like Fable", or proactively when starting any multi-step task that no task-specific skill covers. Subcommands - plan (stop after the plan), audit (grade finished work against the loop), report (rewrite an answer outcome-first).
Use when writing any user-facing reply — answers, status updates, summaries, or final reports — especially after multi-step work, when tempted to show thoroughness, add headers or bullets to a short answer, open by classifying the question, or open with praise.
Convert discovery evidence into bounded, testable work cards with explicit acceptance criteria and architectural invariants. Use when designing multi-file features, planning complex refactors, decomposing architectural migrations, or structuring multi-step implementation tasks — even if the user does not explicitly say "fable-plan" (e.g. "plan this feature", "design the architecture", "break this task down", "create an implementation plan"). Do NOT use when load-bearing facts remain unknown (use fable-discover first) or for trivial single-scope bug fixes (use fable-tdd or fable-execute).
Use before claiming anything works, is fixed, is done, or passes; before status updates on changes you haven't run; before agreeing with someone else's claim about your work; and before state-changing commands like restarts, deletes, or config edits.
Diagnose repeated command failures, stale build caches, branch drift, or contradictory evidence before attempting further code edits. Use when commands fail repeatedly, tests stay red after attempted fixes, build output contradicts source code, or the execution path is confused — even if the user does not explicitly say "fable-recover" (e.g. "still failing", "why did this fail again", "stuck in a failure loop", "diagnose this error"). Do NOT use for routine first-time test failures in fresh TDD (use fable-tdd).
Low-level agentic design patterns and cognitive reflexes across Depth, Breadth, Coil, and Mesh. Turns senior engineering wisdom into reflexes that remove bloat, preserve truth, prevent drift, and compound learning (re0, readchk, autobahn, ssotize, detool, feynman, debloat, sip, hate, macrothink, prism). Use when refactoring complex systems, stripping AI slop, rewriting drifted artifacts from clean v0, consolidating truth into single sources, or evaluating architectural decisions against contrarian counter-arguments.
Retrieves a Jira ticket or a GitHub issue into docs/<KEY>.md as a read-only, validated Markdown snapshot for downstream workflow phases. Use when a Jira URL, a GitHub issue URL, or owner/repo/number coordinates need the Phase 1 fetch-work-item step. Detects the platform from the input and loads the matching playbook just-in-time.
Generate a self-contained JavaScript server runtime and registration metadata for an MCP codeful tool. Use when the user asks to create a codeful MCP tool, generate server logic for an MCP tool, write a runTool function, build a Dataverse-backed MCP tool, or pair MCP server logic with an MCP App widget
The mandatory final handoff after a gm walk reaches phase=COMPLETE with prd_pending_count=0. The agent shall never end a gm session with prose alone, and shall dispatch this skill instead. Searches for remaining work and reloads gm if any exists, or loads wfgy-method if a walk already repeated the same gap. Use immediately when any gm chain reaches its terminal state, never as a substitute for gm itself
Multi-engine long-term memory for game AI agents. Deep Godot 4.7+ (99 skills, 27 genre blueprints) + Bevy docs-first parity — audited patterns, persona orchestration (Analyst · Auditor · Builder), production-grade engineering.
Multi-engine long-term memory for game AI agents. Deep Godot 4.7+ (99 skills, 27 genre blueprints) + Bevy docs-first parity — audited patterns, persona orchestration (Analyst · Auditor · Builder), production-grade engineering.
Keep iterating on code changes until the tests pass, the build succeeds, or linting is clean. Runs in a tight loop of fix → run → check → repeat. Use when you want the agent to autonomously grind through test failures or build errors.
INVOKE THIS SKILL when building evaluation pipelines for LangSmith. Covers three core components: (1) Creating Evaluators - LLM-as-Judge, custom code; (2) Defining Run Functions - how to capture outputs and trajectories from your agent; (3) Running Evaluations - locally with evaluate() or auto-run via LangSmith. Uses the langsmith CLI tool.
Model Context Protocol (MCP) server development and tool management. Languages: Python, TypeScript. Capabilities: build MCP servers, integrate external APIs, discover/execute MCP tools, manage multi-server configs, design agent-centric tools. Actions: create, build, integrate, discover, execute, configure MCP servers/tools. Keywords: MCP, Model Context Protocol, MCP server, MCP tool, stdio transport, SSE transport, tool discovery, resource provider, prompt template, external API integration, Gemini CLI MCP, Claude MCP, agent tools, tool execution, server config. Use when: building MCP servers, integrating external APIs as MCP tools, discovering available MCP tools, executing MCP capabilities, configuring multi-server setups, designing tools for AI agents.
Test and evaluate MCP server tools in the current Claude Code session. Use when auditing MCP configurations, validating tool quality, testing MCP servers end-to-end, generating test cases, checking tool descriptions and schemas, analyzing tool efficiency and redundancy, or debugging MCP integration issues. Covers tool discovery, quality analysis, test generation with AAA pattern, execution, rating, and cross-tool redundancy analysis.
Take a judge live on production traffic — create an Opik online evaluation rule (LLM-as-judge or Python metric) on a project with sampling, filters, variable mapping, and a cost cap, then confirm new traces are being scored. Works over the SDK's REST client; reads rules and score names via the MCP when connected. Returns the rule, the score name it emits, and how to watch it. Use for "score production traces", "monitor hallucinations in prod", "take this judge live", "set up an online evaluation rule", "alert me when quality drops". Not for offline experiments (use evaluate or compare) or for finding what is already broken (use diagnose).
Improve a prompt with the Opik Agent Optimizer — resolve the prompt, a dataset, and a metric, pick the algorithm, run a bounded optimization, check the gain on held-out data, and save the winner as a new prompt version with the optimization run link. Runs via the opik-optimizer package; reads prompts and datasets via the MCP when connected. Use for "optimize this prompt", "improve my system prompt", "make the agent answer better", "tune the prompt against my dataset", "run the prompt optimizer". Not for measuring quality once (use evaluate), before/after on a suite (use compare), or hand-editing a prompt without data.
Meta-skill for auditing and rewriting SKILL.md `description` fields per the agentskills.io optimizing-descriptions framework, layered with mizchi's two-track trigger policy (Meta = explicit-invoke-only, Project = pushy auto-trigger). Invoke ONLY when the user explicitly asks to "optimize a skill description," "audit descriptions," or "rewrite descriptions per agentskills." Do NOT auto-invoke after every SKILL.md edit; description tuning is a deliberate batch, not a per-edit reflex.
Use when user says "paleo mode", "save tokens", "be brief", "terse", "compress output", or says "paleo". Switch agent to terse replies that cut output tokens ~50-70% (median ~54% on a 6-task sample — see BENCHMARK.md) while keeping code, commands, errors, and technical terms byte-exact. Off: "stop paleo" / "normal mode".
Decomposition methodology for pm agent — turns an approved ARCH document into a Beads task list with explicit dependencies, time-boxes, and acceptance criteria. The pipeline can only orchestrate work it can see; this skill defines what "seeable work" looks like.
ALWAYS load before starting any task. Maximizes token efficiency for all file operations, searches, and command execution. Covers extract modes (content, outline, symbols, ast, lines), verbosity tuning, multi-file batching, and discover tool orchestration. Ensures agents spend tokens on outcomes, not overhead.
Manages session state and context handoffs for multi-session projects using the Session Handoff Protocol. Creates and maintains SESSION.md to track phase progress, git checkpoints, and next actions across context clears. Integrates with project-planning skill to convert IMPLEMENTATION_PHASES.md into trackable session state. Use when starting new projects after planning, resuming work after context clear, or managing complex multi-phase implementations. Keywords: session management, SESSION.md, session handoff protocol, context handoff, multi-session projects, phase tracking, git checkpoints, session state tracking, resume work, context clear, phase progress tracking, implementation phases, verification stage, debugging stage, next action tracking, work continuity, session recovery, context management, phased implementation tracking
Track progress across work sessions using SESSION.md with git checkpoints and concrete next actions. Converts IMPLEMENTATION_PHASES.md into trackable session state with phase status, progress markers, and recovery points. Use when: starting projects after planning phase, resuming work after context clears, managing multi-phase implementations, or troubleshooting lost context, missing progress tracking, or unclear next steps.
Defense techniques against prompt injection attacks including direct injection, indirect injection, and jailbreaks - theUse when "prompt injection, jailbreak prevention, input sanitization, llm security, injection attack, security, prompt-injection, llm, owasp, jailbreak, ai-safety" mentioned.
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.