Skills
101,323 packaged capabilities and playbooks for agents. Install any of them with one command.
All
Use when the user asks for Pionex market data: price, ticker, order book depth, recent trades, symbol info (precision/min size), or OHLCV klines. All commands are read-only and do NOT require API credentials. Do NOT use for account balance (pionex-portfolio) or placing/cancelling orders (pionex-trade).
When the user wants to create SEO-driven pages at scale using templates and data. Also use when the user mentions "programmatic SEO," "template pages," "pages at scale," "directory pages," "location pages," "[keyword] + [city] pages," "comparison pages," "integration pages," or "building many pages for SEO." For auditing existing SEO issues, see seo-audit.
Owns Python code style for this stack: ruff for lint + format, numpydoc for docstrings. Three responsibilities — (1) place the project's `ruff.toml` from the bundled template once the stack and workspace are in place, (2) run ruff against any Python files Claude has just generated or edited, and (3) contextualize each touched file's comments to the data-science problem — rewriting any leftover template / workflow prose (skill names, gates, runner, digest, guard-rails) into concise, problem-specific docs so the user's committed files read like a colleague wrote them, not like a generated scaffold. Stops at "the touched files pass `ruff check` and document the problem, not the process." TRIGGER when (any of these): (1) a Python file was just created or edited via Write / Edit / MultiEdit — invoke this skill before declaring the task done so ruff is run AND the file's comments are contextualized to the problem; (2) a fresh ML workspace was just scaffolded by `organize-ml-workspace` and the project has no `ruff.toml` at its root yet — drop the bundled template; (3) the user asks about lint, format, docstring style, or reaches for `black` / `isort` / `flake8` / `pydocstyle` (redirect to ruff — the stack's canonical linter, owned by `data-science-python-stack` Tier 1). SKIP when: the project is non-Python; the only edits in this turn are to Markdown / TOML / JSON / YAML; the file lives in a third-party vendored directory the user doesn't own. HOW TO USE: run ruff manually on the files you just touched — do not configure a PostToolUse hook for this. **Read the "Stop conditions" block and emit the Pre-flight checklist as visible text in your response — both are mandatory before running ruff.**
Review requirements before development. Identify ambiguity, missing information, contradictions, unclear acceptance criteria, logical gaps, edge cases, and testability issues. Determine whether the requirements are sufficiently clear for development and QA. Use this skill when a user provides requirements (BRD, user story, use case, feature ticket, specification, or prose) and asks for feedback, validation, review, or readiness assessment before implementation.
Guides use of the Qdrant Migration Tool CLI to move vectors, metadata, and sparse embeddings from another vector database into Qdrant. Use when someone asks 'how do I migrate from Pinecone/Weaviate/Milvus/Elasticsearch/OpenSearch/pgvector/s3/Azure/DynamoDB/Turbopuffer/Chroma/Redis/MongoDB/FAISS/Solr to Qdrant', 'move my vector database to Qdrant', 'is there a migration tool for Qdrant', 'my migration got interrupted, how do I resume it', 'migration is too slow', or 'collection wasn't created after migration'. Also use when picking migration batch size, deciding on `--net=host`, or verifying data after a migration.
Read GitHub repos the right way — via gitmcp.io instead of raw scraping. Why this beats web search: (1) Semantic search across docs, not just keyword matching, (2) Smart code navigation with accurate file structure — zero hallucinations on repo layout, (3) Proper markdown output optimized for LLMs, not raw HTML/JSON garbage, (4) Aggregates README + /docs + code in one clean interface, (5) Respects rate limits and robots.txt. Stop pasting raw GitHub URLs — use this instead.
Turn a teardown (or a rough idea) into a finished, ready-to-shoot reel. Claude writes the beat-by-beat script in your voice and lays out the on-screen text, the scene for each beat, the pacing cues, and the caption — your words on a proven structure. Use when the user says "build the reel", "make the reel", "write the full reel", "turn this into a reel", "script and storyboard this", or hands over a breakdown/idea and wants the whole thing assembled.
Ship a release: the final pre-tag gate for OrchestKit itself, run once a change is already known good. Walks 12 ordered steps (build, full test suite, security tests, typecheck, manifest count validation, changelog entry, version bump across package.json and CLAUDE.md, selective staging, conventional release commit, tag, confirmed push), halting at the first failed step with remediation guidance. Use when cutting and shipping a version, not for judging whether a change is correct (that is /ork:verify).
Prepare the reproducibility, transparency and open-science parts of a paper, including data and code availability statements, reproducibility checklists (NeurIPS, ICML, ICLR, ACL Responsible NLP, Nature reporting summaries), research artifact packaging (Zenodo DOI, CITATION.cff, environment lockfiles, seeds), ethics/broader-impact statements, CRediT author contributions, competing interests and AI-use disclosures. Use when a venue requires any of these statements or checklists, or when preparing code/data for release alongside a paper.
Performance patterns and gotchas for querying a Rerun catalog from Python. Reach for this when exploring an unfamiliar dataset, when a CatalogClient/dataset query is unexpectedly slow, or when shaping a per-segment / per-episode pipeline that hits the catalog from many places.
Ingest a LeRobot (HuggingFace) dataset into Rerun. Read when converting a LeRobot dataset to RRDs, splitting it into per-episode segments, or registering it on a Rerun catalog. Covers the built-in directory importer (log_file_from_path), the per-episode LeRobotReader, and when to drop to ParquetReader for custom control.
Audits resume ATS parseability, checks JD keyword coverage, and generates ATS-safe output recommendations. Use when users mention ATS, automated screening, keyword matching, resume parsing failures, or job platform formats. Provides explainable risk checks; does not guarantee passing any ATS or getting an interview.
Write and change Rust the way THIS workspace already does it, NOT by generic defaults — a systems-programming reference carrying the real crate boundaries, error model, and conventions for storage engines, binary formats, Arrow/DataFusion, search and vector indexing, async concurrency, testing, and unsafe discipline, so a change lands idiomatic and review-ready instead of merely compiling. Use when a task touches this workspace's Rust crates and should follow its systems-level conventions rather than boilerplate.
Fetch, analyze, reproduce, and minimize GitHub issue reproductions. Use only when the user asks to check whether a GitHub issue reproduces, minimize a GitHub issue's repro, analyze a GitHub bug report, or scrub/triage a GitHub issue for reproducibility. Do not use for standalone reproduction requests without a GitHub issue.
Use when checking your own manuscript before submission from a reviewer's perspective. Returns anticipated Major/Minor comments with fixes, including numerical, citation and leakage checks, with an optional multi-reviewer panel. Someone else's paper is /peer-review.
Use to configure and verify Agentforce agent-to-human escalation, including human handoff, a staffed fallback queue, and failure-threshold directives. Triggers: configure agent escalation, escalate to a human agent, agent handoff to a queue, configure a human escalation queue, set an Agentforce fallback handoff. Do not use to create an agent or configure inbound channel routing; use service-agentforce-channel-configure for inbound routing.
Evidence-based investigative code review using deductive reasoning to determine what actually happened versus what was claimed. Use when verifying implementation claims, investigating bugs, validating fixes, or conducting root cause analysis. Elementary approach to finding truth through systematic observation.
Use this skill for all GitHub write operations in creative-graphic-design organization repositories (such as design-generators), including git commits and pushes, PR creation and updates, issue and PR comments, labels, milestones, and user-requested merges. Perform them as the creative-graphic-design-dev machine user. Use it both when the user explicitly requests the bot or dev account and whenever beginning write work in a creative-graphic-design organization repository without an explicit request; never use it in repositories of any other organization or host.
Use this skill when converting custom PyTorch models to Hugging Face Transformers format. Helps with: (1) Creating PretrainedConfig and PreTrainedModel classes, (2) Writing ImageProcessor/Tokenizer, (3) Compatibility testing, (4) Hub upload preparation. Use when the user wants to make their model compatible with transformers library.
Guidance and workflow for planning, writing, submitting, revising, and publishing scientific manuscripts in high-impact journals. Use when Claude needs to advise on research quality, study design, manuscript structure, journal selection, peer review responses, or publication strategy.
Map who a person knows from one LinkedIn profile URL, when you cannot see their connection list. Derives a per-person angle set from their own work history, education and posts, fans those angles out to parallel agents, merges and scores every candidate on evidence strength, resolves each row to a real profile and current employer, and ships a spreadsheet you can slice by where each person works today (a filterable people tab plus an employer x tier pivot). Works for any function (engineering, recruiting, sales, research, exec) and any career shape. Use for "who does this person know", "find connections of this LinkedIn URL", "map this person's network", "who is in their orbit", "who do they know at a company", "infer connections from job history and post engagement", /social-connections.
Triage and remediate vulnerabilities found by a Strix pentest (open-source CLI or app.strix.ai cloud), then re-run Strix to verify each fix. Use after a Strix scan reports findings, or when the user asks to fix security issues from a strix_runs report, vulnerabilities.json, findings.sarif, or a cloud scan's vulnerabilities.
Root-cause fix discipline for a debugging session already underway. Use when about to propose or apply a fix for a bug or test failure, when tempted to patch a symptom (add a null check, wrap in try-catch, bump a timeout, add a retry), and when a previous fix didn't work. /diagnosing-bugs builds the repro; this governs the fix.
Build view-aligned and screen-space surface effects in Three.js. Use for touch-history frost and thaw, ping-pong accumulation, reduced-resolution blur, crystalline masks, two-scale refraction, and procedural rain droplets that refract and blur a background through wet glass.
Review the execution process, repeated mistakes, and tool waste of the current or specified coding session, or investigate the causes of observed Agent Skill failures and suggest improvements to prevent recurrence. Do not use for general code reviews, static skill catalog audits, or direct implementation.
Activate for the daily repository triage run, or when the user asks to triage CI failures, open issues, or recent commits. Reads yesterday's signals, dedupes against the state file, drafts fixes in isolated worktrees with a reviewer pass, opens PRs on green, and keeps people out of limbo with comments in the maintainer's voice. The state file is the spine: it remembers what was tried, what passed, and what is still open, so each run resumes where the last one stopped.
Unity shaders, materials, and rendering pipelines (URP/HDRP/Built-in). PROACTIVELY activate for: (1) writing shaders in Shader Graph, HLSL, or ShaderLab, (2) URP and HDRP shader authoring, (3) custom render pipeline work (SRP), (4) lighting setup (baked vs realtime, lightmaps, Global Illumination), (5) post-processing stacks, (6) reflection probes and light probes, (7) custom render features and full-screen passes, (8) shader stripping and variant management, (9) compute shaders, (10) ray tracing in HDRP. Provides: Shader Graph templates, HLSL snippets, URP/HDRP differences, lighting setup recipes, render-feature examples, and shader-variant guidance.
After you enter a link from any video site or podcast platform, the corresponding media file is downloaded automatically and delivered to the user. yt-dlp is preferred, covering common video sites such as Douyin, Bilibili, and YouTube, and it can also be used for podcast platforms that expose a direct audio URL (such as a single-episode link from Xiaoyuzhou). When 403 errors, login, age, or region restrictions occur, a retry with cookies.txt is supported; for platforms that may have DRM/encryption or terms restrictions (for example some Spotify content), prompt the user to download only content they have the right to save, and when download is not possible, suggest official offline/export channels or provide the original RSS/direct link. When a Douyin video hits yt-dlp's signature wall ("Fresh cookies needed"), it automatically falls back in three stages: a direct unauthenticated connection to aweme.snssdk.com/v1/play → re-running yt-dlp with local browser guest cookies (no Douyin login needed), all seamlessly; Douyin image-text notes are not yet supported for automatic download and require manual handling.
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.
