Codex Skills

A small collection of reusable custom skills for Codex. Each folder in Skills/ contains one skill and its SKILL.md instructions.

Skill index

Skill	Purpose
agent-optimization-prompt	Audits and rewrites agent prompts, subagent briefs, and instruction files while preserving behavior and authority boundaries.
auteur	Directs explicit cinematic web design and scroll storytelling with art-direction and quality gates.
design-taste-frontend	Researches and builds context-specific landing pages, portfolios, editorial experiences, and brand-preserving redesigns.
debunk	Explicitly fact-checks material claims in a public article or video and traces their evidence and provenance.
debug-loop-recovery	Breaks stalled debugging loops by distinguishing plausible causes and running the smallest evidence-producing next step.
research-synthesis	Produces decision-focused research across relevant primary, repository, academic, industry, and community sources.
skill-regression-evaluator	Explicitly tests skill routing and behavior against realistic positive, negative, boundary, and regression cases.
subagent-team-orchestration	Forms and coordinates a small durable team for authorized, genuinely parallel project work.
visual-asset-direction	Directs consistent image generation and editing through clear reference roles, preservation constraints, and visual review.
visual-plan	Creates evidence-grounded diagrams, impact maps, sequences, or wireframes when a visual materially improves implementation planning.
watch-video	Analyzes public or local videos through bounded scene-aware frames, contact sheets, and available captions.

Upstream credits

Skills that copy or substantially adapt upstream work retain a clear source and license notice:

Skill	Upstream source	License / attribution
auteur	agiwhitelist/auteur	MIT; Copyright (c) 2026 agiwhitelist. The upstream license is included with the skill.
design-taste-frontend	Leonxlnx/taste-skill	MIT; Copyright (c) 2026 Leonxlnx. The upstream license is included with the skill.