> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mirage.strukto.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# SKILL.md

> The Markdown file that tells an agent which CLIs it can run and how to call them.

SKILL.md lists the CLIs installed in a workspace, each with its top-level
help; a subcommand's usage is one `--help` away. It is built from the live
workspace and follows the session's [profile](https://docs.mirage.strukto.ai/home/policy/overview.md): a CLI
or subcommand the profile denies is never mentioned. Each CLI's reference
shows its SKILL.md, such as [Slack](https://docs.mirage.strukto.ai/python/cli/slack.md#skill-md). The mounted
paths are in [VFS.md](https://docs.mirage.strukto.ai/home/setup/vfs-md.md).

## Get it

> [tab] Python
>
> ```python theme={null}
> skill_md = await ws.skill_md()
> agent = await ws.skill_md(session_id="agent")
> preview = await ws.skill_md(profile="reader")
> await ws.skill_md("/SKILL.md")
> ```

> [tab] TypeScript
>
> ```typescript theme={null}
> const skillMd = await ws.skillMd()
> const agent = await ws.skillMd(undefined, { sessionId: 'agent' })
> const preview = await ws.skillMd(undefined, { profile: 'reader' })
> await ws.skillMd('/SKILL.md')
> ```

> [tab] HTTP
>
> ```bash theme={null}
> curl http://127.0.0.1:8765/v1/workspaces/demo/skill-md \
>   -H "Authorization: Bearer $MIRAGE_AUTH_TOKEN"
>
> curl http://127.0.0.1:8765/v1/workspaces/demo/sessions/agent/skill-md \
>   -H "Authorization: Bearer $MIRAGE_AUTH_TOKEN"
>
> curl -X PUT http://127.0.0.1:8765/v1/workspaces/demo/skill-md \
>   -H "Authorization: Bearer $MIRAGE_AUTH_TOKEN" \
>   -H 'Content-Type: application/json' \
>   -d '{"path":"/SKILL.md"}'
> ```

> [tab] CLI
>
> ```bash theme={null}
> mirage workspace skill-md demo
> mirage workspace skill-md demo --session agent
> mirage workspace skill-md demo --path /SKILL.md
> ```

* **No session or profile**: what the workspace's default profile allows.
* **A session**: what that session's agent sees.
* **A profile**: a preview of what a session with that profile would see.
* **A path**: the file appears in the workspace, read-only and always
  current, so the agent runs `cat /SKILL.md`.

## What SKILL.md says about each CLI

* **Usage**: the CLI's top-level `--help`: its groups, options and operands,
  from the same spec its parser uses.
* **Denied subcommands**: when the profile allows only part of a CLI, the
  help pages down to what it allows, listing only the subcommands that lead
  there.
* **Own argument parsing**: a script CLI with no spec shows only its
  registered description.

Mount commands such as `ls` and `grep` are in VFS.md, not here.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.