Loading...
Loading...
Found 10,789 Skills
Enforce allowed and forbidden conversation topics using semantic embedding similarity with session-aware drift detection
Complete subtitle and caption system for FFmpeg 7.1 LTS and 8.0.1 (latest stable, released 2025-11-20). PROACTIVELY activate for: (1) Burning subtitles (hardcoding SRT/ASS/VTT), (2) Adding soft subtitle tracks, (3) Extracting subtitles from video, (4) Subtitle format conversion, (5) Styled captions (font, color, outline, shadow), (6) Subtitle positioning and alignment, (7) CEA-608/708 closed captions, (8) Text overlays with drawtext, (9) Whisper AI automatic transcription (FFmpeg 8.0+ with VAD, multi-language, GPU), (10) Batch subtitle processing. Provides: Format reference tables, styling parameter guide, position alignment charts, Whisper model comparison, VAD configuration, dynamic text examples, accessibility best practices. Ensures: Professional captions with proper styling and accessibility compliance.
This skill should be used when the user wants to bulk-build ArcKit artefacts in parallel rather than running individual /arckit:* commands one at a time. Load whenever the task sounds like 'kick off a build', 'build everything', 'generate all artefacts', 'run all the commands', 'rebuild this project from scratch', 'resume the build', 'pick up where we left off', 'refresh the artefacts', 'run the recipe', 'build the whole project end-to-end', or 'parallel build', or mentions `--plan`, `--resume`, `--target`, `--refresh`, `--recipe`, or `.arckit/state.json`. The skill orchestrates parallel /arckit:* generation using subagent isolation: reads project state, computes the artefact dependency DAG, dispatches one subagent per target per wave (each subagent invokes a /arckit:* skill in its own context), validates outputs, commits the wave, and persists progress to .arckit/state.json for resumability.
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.**
When the user wants to build or improve a sales bot's ability to track conversion rates, drop-off points, and response patterns. Also use when the user mentions "bot analytics," "conversation metrics," "tracking performance," "measuring bot effectiveness," or "conversion tracking."
When the user wants to build or improve a sales bot's ability to introduce scarcity or time-sensitivity without being pushy. Also use when the user mentions "creating urgency," "scarcity," "time-sensitive offers," "limited availability," or "driving action."
Crear o actualizar Architecture Decision Records (ADRs) en docs/adr/. Activar siempre que el usuario quiera documentar, registrar, actualizar o cambiar el estado de una decisión arquitectónica — incluso si no usa la palabra "ADR". Frases que activan este skill: "registrar decisión", "documentar por qué usamos X", "dejar constancia de esta elección técnica", "decision record", "cambiar ADR a Accepted", "marcar como Superseded", "crear ADR", "actualizar ADR", "nuevo ADR", "ADR-XXX". Usar también cuando el usuario describa una tensión arquitectónica que deba quedar documentada.
Datos de fondos comunes de inversion argentinos via CAFCI (Camara Argentina de Fondos Comunes de Inversion). Combina catalogo JSON (1152 fondos, 4615 clases, fees, IDs, metadata), snapshot diario XLSX (VCP, patrimonio, market share, variaciones), ficha individual markdown (rendimientos TNA por periodo) y composicion de cartera (top activos). Sin API key.
Datos de mercado del MAE (Mercado Abierto Electronico de Argentina): renta fija, cauciones, REPO, FOREX, contratos de futuro dolar (DDF), indice ARS-MAE, licitaciones primarias, comunicados institucionales, flujos de fondos para curvas. Sin API key.
OCDNet for scene text detection. Detects arbitrary-oriented text regions in natural images using a differentiable binarization approach. Use when training, evaluating, exporting, pruning, quantizing, retraining, or running inference for a TAO OCDNet model. Trigger phrases include "train OCDNet", "scene text detection", "arbitrary-oriented text boxes", "differentiable binarization detector".
Create a self-contained HTML file for whatever the user is describing, in the effective HTML style. Use when the user wants an HTML artifact that isn't specifically a diagram or a plan — a report, explainer, comparison, deck, prototype, or anything else best delivered as one HTML file.
Shopping price comparison using Bright Data's web scraping infrastructure. Finds where a product is sold, for how much, and whether it's in stock — across Amazon, Walmart, eBay, Best Buy, Google Shopping, and any retailer URL — then ranks the offers into a single buy-recommendation table. Use this skill when the user wants to compare prices, find the cheapest place to buy something, do a price check, see "how much does X cost on Amazon vs Walmart", track an item's price, or decide where to buy a product. Handles product names, ASINs, and direct URLs, and is region-aware (country affects price, availability, and which retailers apply). This is consumer purchase-decision research — for analyzing a competitor's pricing *strategy*, use competitive-intel instead.