Loading...
Loading...
Found 5,478 Skills
Write, push, run, publish, and manage Kaggle Benchmark tasks using the kaggle CLI and the kaggle-benchmarks Python SDK. Use when the user wants to create or push a benchmark task (optionally with attached Kaggle datasets), run benchmarks against LLM models, check task/run status, stream or fetch execution logs, download results and source notebooks, publish a task to make it public, or troubleshoot benchmark workflows.
Structured data research: search sources, extract structured data, archive raw sources, maintain canonical tracker pages, deduplicate. Parameterized via YAML recipes for investor updates, donations, company updates, or any email-to-structured-data pipeline.
The orchestrator and entry point for the engineering skills suite. Use this skill whenever the task involves doing engineering work to a high bar — reviewing code or a design, designing a new system or component, debugging a hard problem or running an incident, implementing a substantive change, writing documentation, or sanity-checking an approach. Use it when the user phrases things casually ("rip into this", "be brutal", "is this approach right", "what am I missing", "what would you change", "look at this") or formally ("review this PR", "audit this design"). Use it proactively for any non-trivial engineering work, before declaring something done. The skill triages the work, dispatches to the right specialty skill(s), enforces verification, and produces an evidence-backed result. The goal is to ensure no AI shortcut, sycophantic agreement, or stylistic distraction gets in the way of work that holds up to senior-engineer scrutiny.
Adds tracing, telemetry, and observability to an assistant-ui backend. Use when wiring an AI SDK route handler (streamText/generateText, toUIMessageStreamResponse) to a tracing backend: Langfuse via OpenTelemetry (LangfuseSpanProcessor and NodeSDK in instrumentation.ts, experimental_telemetry isEnabled, propagateAttributes with traceName/userId/sessionId, langfuseSpanProcessor.forceFlush on serverless), LangSmith via wrapAISDK(ai) from langsmith/experimental/vercel (createLangSmithProviderOptions, awaitPendingTraceBatches), or Helicone via createOpenAI baseURL https://oai.helicone.ai/v1 with the Helicone-Auth header. Also covers rendering collected spans with @assistant-ui/react-o11y headless primitives (SpanResource, SpanPrimitive Root/Indent/CollapseToggle/StatusIndicator/TypeBadge/Name/Children, SpanByIndexProvider, SpanData/SpanState) mounted via useAui/AuiProvider from @assistant-ui/store. Use for missing or empty traces, edge vs nodejs runtime telemetry, serverless flush issues, or trace waterfalls.
Publishing, updating, and serving decentralized websites on Walrus Sites. Use when the user needs to deploy a frontend to Walrus Sites, run a local portal for testnet, debug site-builder errors, configure ws-resources.json, or manage site object lifecycle (update, destroy, extend blobs). Also use when the user asks about site-builder, walrus-sites, portal setup, or hosting a dApp on Walrus. For blob storage without the Sites framework (raw upload/download), see the `accessing-data` skill's walrus.md.
Run and analyze molecular dynamics simulations with OpenMM and MDAnalysis. Set up protein/small molecule systems, define force fields, run energy minimization and production MD, analyze trajectories (RMSD, RMSF, contact maps, free energy surfaces). For structural biology, drug binding, and biophysics.
Manage installation, version tracking, and update checks for Claude Code, Codex, and OpenClaw Skills. Supports installation from local paths or GitHub repositories, automatically identifies .codex/.claude/.openclaw target directories, records installation time, source URL, and version number for each Skill, and checks for GitHub updates.
Owns the smoke test contract for an ML experiment: a small, diagnostic-by-construction pytest that fits the experiment's learner on a portion of the real `data/` source and predicts on a *disjoint* portion that deliberately carries **no pre-history buffer**. The assertion is structural — the number of predictions must equal the number of rows in the predict grid. A pipeline that loads-then-features-then-splits will silently drop the cold-start rows of the predict slice and the test will fail with a row-count mismatch; a pipeline that marks X early and references upstream history nodes from feature steps will pass trivially. The smoke test is the executable proof of the X-marker placement rule from `build-ml-pipeline`. TRIGGER when: `test-ml-pipeline` has dispatched here to write the smoke test for an approved experiment; `pytest tests/smoke/` is failing on row count; the user asks "why is the smoke test failing?"; a pipeline edit in `build-ml-pipeline` needs an executable proof; an experiment script changes the pipeline shape and the matching smoke test needs revisiting. SKIP when: the design note does not exist or is not yet approved (route to `iterate-ml-experiment`); the user is asking about a regression test or schema invariant (route to `regression-test-ml-pipeline` / `distribution-test-ml-pipeline` once those exist); the question is the *interpretation* of CV metrics, not predict-time correctness (route to `evaluate-ml-pipeline`). HOW TO USE: read the matching experiment's `journal/NN_*.md` and `experiments/NN_*.py` first to understand the pipeline's source binding (what env-dict keys does `build_learner` expect?). Then construct two env-dicts from the **real `data/` source** — a train env and a predict env — such that the predict env carries *only the rows we want predictions for* and *no pre-history buffer*. The hard assertion is that the prediction count matches the predict-env row count exactly. The soft assertion is that the smoke set's MAE is within `3 × CV_mean` (or the task-appropriate analogue). **Do not write the design note or run CV — that's other skills' job.**
Verify citations and references in scientific documents to detect hallucinated or invalid sources. Extracts DOIs, URLs, arXiv IDs, PubMed IDs, and ISBNs from Markdown, LaTeX, org-mode, and plain text, then validates them using API lookups and web fetches. Use this skill when: - Reviewing AI-generated content for citation accuracy - Validating references in papers, reports, or documentation - Checking if DOIs/URLs resolve to actual papers - Auditing a document for broken or fake citations
Adversarial due-diligence on a benchmark you envy — a founder, KOL, company, or product whose claimed success you suspect is inflated. Inline four-phase orchestration — fan-out collection, adversarial verification grading every claim L1-L4 to split marketing bubble from real signal, attribution weighting (product vs timing vs IP vs luck, what's replicable), then mapping the validated playbook onto the user's own resources. Use whenever the user wants to 尽调/对标/拆解 a competitor or role-model, 抄/偷师 someone's playbook, suspects 水分/泡沫 in their claims (Product Hunt
Edit a vigiles .spec.ts to change a compiled instruction file (CLAUDE.md / AGENTS.md) — add, modify, or remove a rule, section, command, or key file. Use whenever you need to change a CLAUDE.md/AGENTS.md that carries a vigiles hash (edit the spec, never the artifact), including adding a new enforce()/check()/guidance() rule.
Guided discovery of a product idea by mining what the founder already knows or already does — covers source selection (business vs. expertise), context capture, pattern synthesis, candidate scorecard, and writes `docs/product-idea.md`. Use when the founder says "generate an idea", "help me find an idea", "what should I build", "product idea from my business", "product idea from my expertise", or otherwise needs to discover a product concept worth building.