Watch Skill gives coding agents a video input: it extracts frames, reads on-screen text, transcribes speech and indexes the results for follow-up questions. The author shares the open-source package and its iterative THE LOOP workflow.
Communauté Agent Skills
Explorez des publications vérifiées sur X, les personnes qui les partagent et les projets qu’elles mentionnent.
Publications par jour
29 publications recenséesSeules les publications sur les Skills recensées et validées sont comptées. Une journée vide signifie qu’aucune publication n’a été recensée ici. Dates en UTC.
Publications récentes
29 publications recensées
The author shares an open-source AI film crew with 11 Skills, from screenplay and visual design through sound and final editing. A 26-second film illustrates the workflow in Claude Code and Codex; the post emphasizes the recorded production decisions as well as the finished film.
Cinetic guides coding agents through making launch videos and short films from code. The creator shares a video made with a short prompt and links the open-source Skill.
The author shares douyin-download for fetching one Douyin work from share text, a link or an ID. It can save the available original images, clean video stream, Live Photo media and BGM, with metadata and SHA256 checks. It requires a logged-in browser and does not collect creator profiles in bulk.
Packages a character-replacement video workflow, from reference-video checks and reusable character profiles to prompts and comparison clips.
This share highlights qisi-video-remix, which adapts a reference video into a new script, character and scene plans, storyboards, and copyable prompts for each video segment. The repository’s default deliverable is a production plan.
Anime Reaction GIF makes looping anime reaction stickers from generated character poses and timed holds. The creator shares a catgirl sneeze example and the open-source Codex Skill.
Pratzyy shares dissect, a Claude Code Skill that measures a reference video and writes a reusable production blueprint. His Kurzgesagt comparison shows an original short beside a remake made from the blueprint; the Skill itself produces the blueprint, rather than rendering the remake.
The author shares six Douyin transcript workflows: single videos, three-video trials, full archives, resuming, metric updates and separate proofreading.
Guizang describes a daily AI news video workflow in Muse: curate notes and media, then use Pi with K3 and GLM 5.3 to assemble a video with guizang-product-video-skill and return it in chat. This is his own integration example.
The creator shares Zuse’s launch-video Skill in a same-author reply. The workflow combines HyperFrames, accurate Zuse UI, Blender assets and music timing for continuous product and partner-announcement films.
Alexey Fateev shares lyric-music-video with a video attachment. The linked Skill turns a finished song and lyrics into word-timed typography and animated scenes using code and generated still images, with a style choice and full-section review before the rest of the song is built.
Build Fast with AI demonstrates a Reddit-themed motion graphics brief with Snoo, kinetic typography and a feed montage, and links its Motion Studio Skill in an author reply.
ACΞ recommends brag for turning a finished project into a short launch video with music and share copy. The post also points to brag-slim, the lean Opus 5.5 route that does not use Hyperframes.
恒星sun recommends onetake’s approach to product motion clips: plan which element carries each beat into the next scene, then check continuity in the rendered frames. The source uses a noncommercial license.
AH Douyin archives a creator’s authorized public videos into local machine transcripts, timed subtitles and a searchable collection. This v0.4.0 post describes profile-based batch collection and sorting an existing archive by likes. The repository uses a source-available non-commercial license; commercial use requires written authorization.
A recommendation of fframes and its video Skill: Rust and SVG scenes, GPU rendering, and an inspection loop using frame images, contact sheets and audio measurements.
Hengxing Sun recommends Guizang’s product-video Skill for turning code and release changes into editable promotional video projects. The post emphasizes real product components, design consistency, timeline animation and sound, with component and dependency integration as prerequisites.
Huangbai shows an AI-made tutorial video and says he guided it through a few conversations. A follow-up reply links the related GitHub repository. SkillsMP has not independently reproduced the result.
aichan1224news recommends onetake for product launch clips, teasers and feature demos. The post focuses on carrying visual elements between scenes and checking continuity with scripts. The project uses a noncommercial license.