Generates a complete 30-40s vertical "A vs B" comparison video project with fixed 3-zone layout and Vietnamese TTS voiceover.
Generates a complete 30-40s vertical "A vs B" explainer video project with Vietnamese TTS voiceover and audio-accurate caption timing.
A step-by-step playbook for designing, implementing, hardening, and verifying production-grade MCP servers.
A skill for designing reproducible AI agent evaluations with representative datasets, rubrics, graders, regression gates, and a release decision memo.
Designs and verifies auditable human approval gates, escalation paths, and safe state transitions for risky AI agent actions.
A skill for designing privacy-aware traces, spans, metrics, alerts, and investigation runbooks for AI agents.
The designer lane of codeck: reads deck.md, defines a validated design system, and assembles a single self-contained HTML presentation.
Generates React/Tailwind v4 components that strictly follow StyleSeed's registry, tokens, and accessibility rules.
Builds or rebuilds a UI screen using StyleSeed's composed rule set, then forces a code score gate and a pixel verification gate before reporting results.
The official operating manual that teaches Claude to build widgets and views for a moi web workspace, plus theming, env handling and verification.
Turns a novel's source bible and product brief into three genuinely different game concepts, then picks the most worthwhile prototype via hard vetoes and explicit trade-offs.
Compresses a novel into a citation-backed SOURCE_BIBLE of game-relevant facts, without picking a genre yet.
Launches the real build on its target runtime and proves six gates — launch, render, input, core loop, a designed outcome, and restart — recording everything in qa/verification.json.
Turns an approved game design into a minimal BUILD_BRIEF plus a runnable candidate, iterated with replayable evidence.
Phase 1 of the prd-taskmaster pipeline: brainstorm-driven discovery that extracts constraints, classifies project scale, and hands off to spec generation.
Audits a paper (PDF/DOI/arXiv) across image, numerical, and logical evidence tracks and produces a Level 1–4 graded, reviewable evidence report.
Turns a research question into a complete experiment package: design doc, runnable code, results.json with provenance, publication-grade figures, and a structured report.
A codeArbiter skill that scans an existing codebase with six parallel read-only scouts and back-fills the full set of project-state documents without guessing.
A strict commit-discipline skill that lets nothing reach version control until nine blocking gates — from explicit authorization to a clean post-commit tree — are cleared.
A bounded multi-agent command protocol that splits genuinely independent work across 5–100+ isolated workers, verifies outputs independently, and integrates serially.