All skills

Turns raw sources dropped into a vault into source, concept and entity pages that are linked in both directions, with index and log updated in the same run.

A read-only skill that audits an agent's context against four places — system prompt, tools, history, tail — and ranks fixes by tokens saved per turn.

A read-only audit of your agent's harness across containment, guides, sensors and permissions that names the worst possible session and ranks hardening fixes outside-in.

Imports a large archive into your vault in oldest-first batches of ten, checkpointing and committing after each batch while keeping cost visible.

Turns a vague task into a machine-checkable definition of done, writes goal-test.sh, and optionally wraps a headless agent in a bounded retry loop.

Audits a pipeline for bounded decisions that don't need a frontier model and proposes rule / classifier / small-model gates with fail-closed escalation.

Turns an exported Claude/ChatGPT history into a privacy-reviewed, triaged list of what to ingest, archive, or delete.

An 11-step workflow that brainstorms brandable .com names and verifies availability, aftermarket price, trademarks and social handles.

Learn a codebase by re-deriving its design decisions from first principles through Socratic questioning instead of reading explanations.

A positioning playbook for AI products accused of being "just a GPT wrapper," helping you name your real moat and answer the two objections you'll always get.

Turns your founder brief and market research into an honest diagnosis plus a stage-aware Now/Next/Later GTM roadmap saved to disk.

Forces the agent to answer Meta ads questions only from reads made in the current turn, always stamped with the data window, sync time and connection state.

Transcribes the finished video's own audio and burns word-timed captions into vertical video ads.

Generate a single photoreal or designed image with OpenAI gpt-image-1/gpt-image-2 through fal.ai, with optional reference-image editing.

Faithfully archives public WPS/KDocs/ProcessOn links as raw API data, original SVG/PNG, and structural Markdown without logging in.

A step-by-step playbook for diagnosing the Scrapling CLI, choosing static vs. browser fetching, and validating extracted output.

Turns reference UI screenshots into a documented design system, an MVP PRD, and an implementation-ready UI prompt for React + Tailwind.

Uses your logged-in Chrome session to pull full-size images out of a Google Gemini conversation, rename them in order, and bundle them into a verified ZIP.

Sets up a full QA process — AAA test cases, P0–P4 bug triage, quality-gate metrics, and OWASP security coverage — with one init script.

A wrapper skill that installs the official Tencent IMA skill in one command, diagnoses and repairs known loading bugs with consent, and runs priority-aware fan-out search across your knowledge bases.