By media people, for media people
mediaskills is open source because the media community deserves agent tooling that speaks our language — timecode, loudness, captions, deliverables, and the messy reality of real files.
We're editors, engineers, and operators who've lived inside NLEs, ffmpeg logs, caption pipelines, and delivery specs. mediaskills isn't a generic "AI wrapper" — it's practical infrastructure: deterministic scripts, stable operation IDs, and cookbooks that mirror how media work actually gets done.
Every skill is designed so an agent can probe before it cuts, chain outputs between steps, and know when not to use a tool. That's the bar we hold ourselves to.
Assistants are becoming part of the edit bay, the QC pass, and the delivery checklist. But most agent tooling still treats video like any other file — no drop-frame math, no caption compliance heuristics, no instinct for when to copy streams vs re-encode.
mediaskills is a first step. It packages media literacy into portableAgent Skills that any compatible agent can install: inspect a master, trim on timecode, normalize a podcast, validate captions, export SCC. The catalog will grow with the community — this is the foundation, not the ceiling.
13 skills, 80+ operations, cookbooks for common delivery and QC paths.
More broadcast formats, deeper vision workflows, and skills you help define.
mediaskills is MIT-licensed and maintained in the open attimelapsetech/mediaskills. We built it to support editors, engineers, post houses, indies, and anyone automating media with agents — not to lock capabilities behind a walled garden.
If mediaskills saves you time on a deliverable, teaches an agent something useful, or sparks an idea for a skill we should add — we want to hear from you.
Time Lapse Technologies backs mediaskills.ai and the open-source project so the community can rely on them long term. This site and the skills catalog will stay free forever — no paywalls, no gated features on the open catalog.
Found a bug? Need a skill we don't have? Open an issue or send a PR. No contribution is too small — doc fixes, recipes, and new scripts all matter.
Something broken? Wrong output, missing binary, or docs that don't match behavior.
Need a workflow or operation we don't cover yet? Tell us what your agents struggle with.
Help agents pick the right skill — when-to-use boundaries, cookbooks, or AGENTS.md.
Pull requests welcome. See CONTRIBUTING.md for smoke tests and skill conventions.