> ## Documentation Index > Fetch the complete documentation index at: https://docs.eversince.ai/llms.txt > Use this file to discover all available pages before exploring further. # Skills > A skill is a saved method an agent reads when the job is that skill's kind. Eversince's own skills carry the craft of editing, design and generation. A workspace's skills belong to its users, or were saved for them by an agent. ## Reading skills `manage_skills` with `list` returns every skill this user can use in this workspace: its id, name, and one line saying what it makes and when it applies. `get` returns one skill in full, or several by id, such as a skill and its parts in one call. Nothing is loaded into the prompt in advance. Open a skill when the work is that skill's kind. A skill body is at most 25,000 tokens, the lowest default a client passes to an agent whole. Over MCP the read declares its size, so a client that truncates results keeps it whole. A batch read that would pass the ceiling stops and names `next_ids`. ## Eversince's skills Each skill holds what one kind of job needs, from the material to the tool calls in order, and each states that a brief, a reference or a brand the user gives outranks its defaults. **Editing** | Skill | What it covers | | ----------------- | ------------------------------------------------------------------------------------------------- | | Talking head | One person to camera, cut to hold a viewer: takes, fillers, punch-ins, B-roll, chapters, end card | | Rough cut | Long raw material built into one piece whose structure can be judged | | Multicam | A conversation recorded on several cameras, the shot following whoever is talking | | Clipping | The moments in something long, each cut to stand alone as a set of clips | | Repurpose | A finished piece in another ratio, length, platform or subtitle language | | Cutouts and masks | A person or subject separated from the background, and the shots built from the two halves | | Color grade | Shots corrected to one reference, then one look over the piece | | Captions | Caption lines cut from the speech, timed, sized and legible over any footage | | Sound design | Music, voice, effects, silence and the mix, and cutting picture to music | | Localization | A finished piece in another language, as a dub in the same voices or as subtitles | | Screen demo | A screen recording turned into a demo, with dead time cut and each step framed | | Vlog | A day lived on camera, cut as a cold open, a stamp and scenes held long in the day's order | | Glasses footage | Clips from camera glasses cut into a short first-person video | | Reference match | A piece built to match a reference and compared with it at the same seconds | | Finish | The last pass over a whole piece before it leaves | **Formats and ads** | Skill | What it covers | | -------------- | ---------------------------------------------------------------------------------- | | Social short | A vertical short around one idea: hook, shape, beats and cut | | UGC ad | A video ad that reads as a real person filming, from real takes or generated | | Static ad | A still built to sell: headline, product and layout, sized for a placement | | Performance ad | A set of ads run against spend, each variant testing a stated question | | Explainer | A narrated video that makes one idea understood | | Launch video | A narrated case for a product, release or pull request, with real screens as proof | | Product film | A film where the product is the hero and the picture carries it | | Trailer | A release made wanted without being explained, landing on a title and a date | | Music video | Pictures, footage or type cut to a track, with the music setting every cut | **Stills and design** | Skill | What it covers | | ---------------------- | ------------------------------------------------------------------------------------------ | | Design | Type, color and composition for anything made to be seen, with ten named looks as its part | | Thumbnails and posters | The one image that represents a piece of content, judged at the size it runs | | Product stills | AI-generated product photography from a real image of the product | | Photo editing | Corrections to a photograph that exists: light, color, crop, background, local fixes | | Slides and carousels | A deck for a room or a carousel for a feed, one claim and one visual per slide | **AI generation** | Skill | What it covers | | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------ | | Cinema | AI-generated video designed and prompted like film | | Animation | AI-generated animation, one style held across shots | | Consistent character | An AI-generated character kept identical across every later generation | | Motion graphics | Graphics built as code on the timeline: titles, lower thirds, charts, kinetic type, with planning, overlays and recipes as its parts | **Research and reference** | Skill | What it covers | | ------------------- | ------------------------------------------------------------------------------------------------------- | | Social research | Posts and ads pulled from the social platforms and ad libraries, read for angles, hooks and buyer words | | Timeline operations | The reference for every `edit_timeline` operation: fields, units, defaults, effects, keyframes | | Canvas operations | The reference for every `edit_canvas` operation: layer kinds, boxes, masks, grades, slides | Eversince's skills can be switched off for one user. They cannot be edited or deleted. ## A workspace's skills `manage_skills` creates, updates and deletes a workspace's own skills. They belong to the workspace: `update` replaces the whole body for everyone, and `delete` reports how many colleagues still use the skill. A skill an agent saved is marked as such. Switching a skill off is per user, and an agent connected as that user does not see it on any interface. A skill body is text. Model names, prices and tool lists written into one are not updated by the system.