Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
All
Use when substituting a real collaborator with a test double — phrases like "mock object", "test double", "stub", "fake", "spy", "Mockito", "MockK", "should I mock this", "database in tests", "fake the time", "fake the HTTP client". Defines test doubles in their five flavors (dummy / fake / stub / spy / mock) with picking criteria, and the strong warnings about over-mocking.
Use when establishing TDD as a discipline — phrases like "red green refactor", "TDD cycle", "test first", "should I do TDD here", "TDD discipline", "what's the TDD loop", "when should I refactor". Defines the canonical TDD cycle (red → green → refactor), what each phase actually requires, the honest scope of when TDD applies, and the cases where it shouldn't.
Creates high-trust, high-performance team environments through structural mechanisms for psychological safety, shared consciousness, and empowered execution. Use when building team culture, onboarding new teams, diagnosing team dysfunction, or improving collaboration.
TeamLead orchestration — use when a Human wants coordinated implementation/delegation for product code, tests, config, infra, features, bug fixes, refactoring, or behavior/UI/API changes. Do not trigger for read-only inspection/status/explanation/lightweight summary/no-edit review/research unless the Human explicitly requests TeamLead orchestration or delegation would clearly help. Work that creates or modifies behavior must pass the Research Gate (spawn research agents immediately after classification) and always invoke `superpowers:brainstorming` first, then the Lead performs the Spec/Plan gates before delegating; the Lead must not write product code themselves. Every spawn must pick a spawn mode (one-shot Agent vs teammate) and specify the model per the Model Policy (Fable only for thinking/planning work, Opus for Dev/QA/research). Execute in batches per the Commit Plan, run tests per the fast/heavy policy (full suite + E2E only once at the end), Human checkpoints only at 3 points, and every prompt includes the repo's Convention Anchor.
Use personal preferences and project evidence to make trade-offs on tech stack, frameworks, tools, data formats or repository boundaries, and set up the Python toolchain (uv, ruff, ty, pyproject, prek/CI). Use when the main outcome is "what to choose, why it fits, when to deviate from the baseline", or when the user asks to create/migrate Python engineering configuration. Do not trigger merely because the request contains "architecture" or "refactoring": the project's next step, module/interface/state boundaries and multi-phase rollout are the responsibility of pilot; this skill only handles the technical choices that genuinely exist within them.
Write dialog test specs for a Telegram bot. Covers why tokenless testing exists, how the harness replays Updates and captures API calls, BotSpec JSON format, coverage rules, and the test harness CLI. Tests are the objective publish gate — all specs must pass for the bot to publish. USE FOR: write bot tests, dialog specs, harness, command coverage, BotSpec JSON, tests/specs/*.json, tests/commands/*.json, GATE verdict, coverage gate — even if the user doesn't say "specs" or "test" explicitly. DO NOT USE FOR: programmatic vitest tests with mocks and edge-case fixtures (see telegram-test-advanced), or session persistence (see telegram-bot-sessions). Triggers: write bot tests, dialog specs, harness, command coverage, BotSpec JSON, tests/specs/*.json, tests/commands/*.json, GATE verdict, coverage gate.
Design, build, test, and distribute external Temps plugins with TypeScript/Bun; provide development and local-testing guidance for existing Rust plugins. Use for creating a plugin, adding an embedded console UI or deployment-event automation, testing host permissions and AI, installing from GitHub, or submitting a plugin to the public catalog. End-to-end publishing covers TypeScript GitHub-source plugins, not Rust native-package releases or in-process TempsPlugin backend crates.
Create distinctive, production-grade terminal user interfaces with high design quality. Use this skill when the user asks to build CLI tools, TUI applications, or terminal-based interfaces. Generates creative, polished code that avoids generic terminal aesthetics.
Create a Terse workflow. Use when the user wants to build an automation that reacts to events (GitHub PRs, Slack messages, Linear issues, cron schedules, webhooks) and takes actions, or wants to get started with Terse. Bootstraps a new Terse project first when none exists.
Optimize unit, integration and end-to-end tests for type selection, assertions, mocks, test data and review quality. Use when the user asks to add, fix, refactor or review tests, or deals with meaningless assertions, mocks, coverage, spec, integration or e2e boundaries.
Govern flaky tests and suite reliability: rerun-pass verdicts, root-cause classes, quarantine, retry semantics, health metrics. Not for: in-run failures (e2e/api), triage, confirmed bugs. 治理 flaky 测试与套件可靠性：时好时坏/重跑变绿判定、根因四分类、隔离门禁、重试诚实语义、健康度。不用于：执行中单条失败（e2e/api）、批量分流（triage）、已确认 Bug（bug-analysis）。
Use when writing or reviewing Foundry tests, deciding whether a suite is enough to deploy, or when a green run or a coverage number is offered as proof that a contract is safe. Covers fuzz, fork and invariant testing. Not for auditing source by reading it (`/security`, `/audit`).
Use when creating or editing skills, before deployment, to verify they work under pressure and resist rationalization. Applies the RED-GREEN-REFACTOR cycle to process documentation by running a baseline without the skill, writing to address failures, and iterating to close loopholes.
The single front door to the SDD pipeline (docs/plans/): research → plan → workshop → ADR → phase tasks → implement → progress → review → ship. Use when the user wants to plan, research, explore, specify or architect (now one atomic 'plan' step → one document: business spec + implementation plan), clarify, workshop, write an ADR, break a phase into tasks, implement or build a phase (optionally with a live review companion), update progress, code-review, or ship a plan (push + PR + watch checks) — or types /the-flow (or a legacy /plan-N command), or asks to start, resume, or adopt a plan flow. Guided mode (no args) coaches from durable on-disk state; direct jump runs one stage: /the-flow <id> <verb> [flags] — 1a explore, 1b plan (spec + plan in one doc; incl. clarify), 2c workshop, 3a adr, 5 tasks, 6 implement (--companion for live review), 6a progress, 7 review, 8 ship, 8c reconcile. Ids and verbs resolve alone; printed commands carry both.
Verify whether an AI agent's decision is well-reasoned before executing. Adversarial multi-model critique (Claude, Grok, DeepSeek) returns ALLOW, BLOCK, or UNCERTAIN with confidence score. Use before any high-value swap, trade, or autonomous action.
Researches public Threads (Meta) profiles, posts, replies, and search results via the Crawlora API, returning clean JSON. Use when the user wants a Threads profile's stats, a post's content/replies, a profile's recent posts, or a keyword search on Threads — instead of scraping the app.
Research public TikTok profiles, videos, comments, hashtags, search results, trending feeds, and Creative Center signals through Crawlora. Use for TikTok-specific discovery or verification instead of scraping TikTok directly; use tiktok-ad-research for ad-only comparisons and influencer-discovery for cross-platform creator shortlists.
Inspect and edit open Tiled maps through the tiled-ai MCP server, including image-only TSX creation, Wang terrain sets, terrain painting, visual tile selection, structures, objects, properties and collisions. Use to set up the integration or build and modify content directly in Tiled.
Add a new cuTile GPU kernel operator to TileGym. Covers dispatch registration in ops.py, cuTile backend implementation, __init__.py exports, test creation, and benchmark in tests/benchmark. Use when adding, creating, or implementing a new cuTile operator/kernel in TileGym, or when asking how to register a new cuTile op.
Converts cuTile Python GPU kernels (@ct.kernel) to cuTile.jl Julia equivalents. Handles kernel syntax translation, 0-indexed to 1-indexed conversion, broadcasting differences, memory layout (row-major to column-major), type system mapping, and launch API differences. Use when converting, porting, or translating cuTile Python kernels to Julia cuTile.jl, or debugging/optimizing existing Julia cuTile translations.
Converts cuTile GPU kernels (@ct.kernel) to Triton (@triton.jit). Handles standard in-repo conversion, debugging (cudaErrorIllegalAddress, shape mismatch, numerical mismatch), and mapping cuTile idioms (ct.load/ct.store, ct.Constant, ct.launch) to Triton equivalents. Covers dual-kernel layout flags (e.g. transpose=True/False + autotune grid via META) per translations/advanced-patterns.md. Use when converting, porting, or translating cuTile kernels to Triton, or debugging existing Triton translations
Use when adding, modifying, optimizing, or debugging CuTile autotuning code. Trigger signals: `exhaustive_search` / `replace_hints` / `hints_fn` / `cuda.tile.tune` in code, `autotune` in filenames, or correctness/performance issues in autotuned CuTile kernels. Covers: tune-once/cache/launch pattern, per-architecture configs (sm80–sm120), parameter space design (tile sizes, occupancy, num_ctas), and 7 common pitfalls with solutions.
Iteratively optimize cuTile kernel performance through systematic profiling, bottleneck analysis, IR comparison, and targeted tuning. Covers tile sizes, occupancy, autotune configs, TMA, latency hints, persistent scheduling, num_ctas, flush_to_zero, and IR-level debugging. Use when asked to "optimize cutile kernel", "improve kernel perf", "tune cutile performance", "make kernel faster", or iteratively benchmark and refine a cuTile GPU kernel in the TileGym project.
Integrate TileGym kernels into Hugging Face `transformers` models by replacing the library's submodule(s) and certain class(es)' implementations, and patching certain class(es)' init/forward/load weight methods prior to instantiating models. Used when the user requires integrating TileGym kernels into `transformers` models
This skill should be used when the user asks to "write a Tiltfile", "configure Tilt", "set up live update", "debug Tilt", "add a resource to Tilt", "optimize Tilt builds", "view Tilt logs", "restart a Tilt resource", or mentions Tiltfile, tilt up, tilt ci, tilt down, live_update, docker_build, custom_build, k8s_resource, local_resource, or Kubernetes local development with Tilt.
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.