Loading...
Loading...
Found 1,429 Skills
Use this skill to turn a raw user request into a structured, model-agnostic task brief before execution, and — when the brief survives confirmation — to be the sole entry point that creates a task directory at `.agents/tasks/<task-id>/`. Invoke whenever the request is complex, multi-step, cross-domain, ambiguous, or will be handed off to another model or agent. Also trigger when the user says things like 'help me figure out what I need', 'I'm not sure how to ask this', 'I want to do X but I don't know where to start', 'take this and make it clearer', or when the task mixes multiple goals or domains. Do NOT trigger for simple one-line requests with clear intent (e.g., 'fix the typo on line 42', 'rename this variable').
Produces a topic-segmented post-meeting summary for attendees with decisions highlighted and actions captured inline per topic (plus a consolidated action view at the end). Auto-populates topic skeleton from a sibling meeting-agenda when available and reconciles planned vs. actual topics. Accepts transcripts from Zoom, Meet, Otter, Fireflies, Krisp MCP, or manual notes; runs on variable-quality input without blocking.
Embed Omni Analytics dashboards in external applications — URL signing, custom themes, iframe events, entity workspaces, and permission-aware content — using the @omni-co/embed SDK and Omni CLI. Use this skill whenever someone wants to embed a dashboard, sign an embed URL, customize the embedded theme, handle embed events, listen for clicks or drills in the iframe, send filters to an embedded dashboard, set up entity workspaces, look up embed users, build a permission-aware content list, white-label an embedded dashboard, or any variant of "embed this dashboard", "customize the iframe theme", "handle click events from the embed", "filter the embedded dashboard", "set up embedding", or "what dashboards can this user see".
Async media + document derivations via `platform.media.transforms` and the declarative `transforms` block in `maravilla.config.ts`. Media: transcode video, thumbnail extraction, image resize/variants, OCR. Documents (.docx/.odt/.pptx/.xlsx/...): convert to PDF, render page thumbnails, generic format conversion, Markdown extraction (RAG-ready), single-file HTML with inlined images, image-replacement templating ({{TAG}} swap + named-object swap), QR-code injection. Use when ingesting user uploads that need normalised renditions, generating contracts/invoices from templates, or extracting structured content for LLMs. Critical: derived keys are content-addressed — `keyFor(srcKey, spec)` is known up front, before the worker starts, so clients can render placeholder UI without round-trips. Declarative config is the default; imperative `transforms.*` calls are for one-offs.
This skill should be used when the user wants to review code, audit a diff, get a second opinion on changes, or run an adversarial review of files in the current working tree. Common triggers include "review this code", "audit this diff", "find issues in", "second opinion on this", "harsh review of", "adversarial review", and "security review of". Picks one or more reviewer personas (adversarial, security, architecture, performance). Reviews local files, `git diff`, or `git diff --staged` only — does not fetch external content. Runs in one of four modes: single-agent (one persona in the current agent), cross-model handoff (independent second opinion via another local AI CLI, with secret-shield preflight + prompt-shield wrap), multi-bg-agent (one persona per parallel background subagent), or agent-team (Claude Code Teams or equivalent on supporting agents). Skip when the user wants formatting fixes (use a linter) or refactoring patterns (use ts-best-practices or ts-best-practices-functional).
Cross-functional what-if modeling for cascading multi-variable scenarios. Unlike single-assumption stress testing, this models compound adversity across all business functions simultaneously. Use when facing complex risk scenarios, strategic decisions with major downside, or when the user asks 'what if X AND Y both happen?'
Design and build UIs with Google Stitch using @_davideast/stitch-mcp as the proxy and CLI. Covers screen generation, editing, variants, design systems, site building, prompt engineering, and the DESIGN.md spec. Updated for Stitch's March 2026 infinite canvas, 4-mode AI, and first-class design systems.
Disciplined spec-driven test-driven development workflow for building software with AI coding agents. Transforms ambiguous requests into verified implementations through structured specification, test derivation, and strict TDD. Handles greenfield projects, brownfield enhancements (with or without existing tests), refactors, and complex bug fixes with workflow-specific guidance for each. Use when the user requests a new feature, module, enhancement, refactor, API, data pipeline, CLI tool, or system with multiple requirements, edge cases, or unclear specifications. Also use for complex bug fixes requiring root cause analysis. Triggers on phrases like "add a feature", "implement", "build a new module", "build an API", "build a CLI", "build a data pipeline", "refactor", "fix this bug", "write tests for", "TDD", "test-first", "the requirements are unclear", "characterization tests", or "spec this out". Triggers when modifying code with adjacent test files (`tests/`, `*_test.py`, `*.test.ts`, `*.spec.ts`, `spec/`, `__tests__/`) or test framework config (pytest.ini, jest.config.*, go.mod with testing imports, Cargo.toml with [dev-dependencies], package.json with a test script). Triggers when the user mentions edge cases, invariants, acceptance criteria, EARS notation, or red-green-refactor. Do NOT use for simple one-line fixes, cosmetic changes, formatting, renames, dependency bumps, or tasks where requirements are already fully specified with tests provided.
Mid-conversation reflection skill that pauses execution and zooms out from detail-mode to honestly reassess direction, assumptions, and bias. Use when the user says 'reflect', 'take a step back', 'step back', 'zoom out', 'are we missing something', 'bigger picture', 'sanity check this', 'are we on track', 'are we overthinking this', 'forest for the trees', or any variation signaling intent to break out of detail-mode and reassess. Also trigger when the conversation has gone deep on implementation details without strategic check-in, or when the user shows signs of being stuck — that's often a signal the framing needs a reset, not more detail work. Intentionally low-intake: runs the 5-dimension analysis immediately when prior context is rich enough; asks one forcing clarifier only when invocation context is too thin to reassess from.
Use when writing or refactoring proof-carrying code in MoonBit, especially for Why3-backed specifications, abstraction functions, representation invariants, proof assertions, recursive verified data structures, or reducing trusted proof bridges.
Implement Thompson sampling for multi-armed and contextual bandits. Use when the user wants to adaptively allocate traffic across variants (ads, recommendations, content, pricing) to minimize regret instead of running a fixed-allocation A/B test. Covers Bernoulli bandits, contextual bandits, regret analysis, and comparison with epsilon-greedy and UCB.
Nature figure preparation: resolution (300+ DPI), formats (AI/EPS/TIFF), RGB color, Helvetica/Arial fonts, lowercase panel labels, image integrity requirements.