Agent Skills


ID	`skills`
Category	Skills
Features	None
Dependencies	`session_file_system`

Discover and activate skills from /.agents/skills/ in the session filesystem. Skills are portable instruction packages following the Agent Skills open specification.

Tools

`list_skills`

Scan /.agents/skills/ for available skills. Returns names and descriptions only (~100 tokens per skill).

`activate_skill`

Load a skill’s full instructions by name.

Parameter	Type	Required	Description
`name`	string	yes	Skill name (directory name under `/.agents/skills/`)

Returns: full SKILL.md content and list of bundled files.

How It Works

Skills use progressive disclosure to keep context efficient:

Discovery (~100 tokens) — list_skills returns only names and descriptions
Activation (<5000 tokens) — activate_skill loads the full SKILL.md instructions
Resources (on-demand) — bundled files accessible via File System tools

Use Cases

Project-specific workflows — upload skills for your project’s deployment, testing, or review processes
Specialized knowledge — domain-specific instructions (e.g., security review checklists)
Reusable templates — code generation patterns, documentation templates

Example

Skills are uploaded to the session workspace:

/.agents/skills/
  deploy/
    SKILL.md          # Deployment instructions
    templates/
      k8s-deploy.yaml # Kubernetes template
  code-review/
    SKILL.md          # Review checklist and process

Agent discovers and uses them:

Agent:
  → list_skills()
  ← [{ name: "deploy", description: "Production deployment workflow" },
     { name: "code-review", description: "Code review checklist" }]

  → activate_skill({ name: "deploy" })
  ← { instructions: "## Deployment Process\n1. Run tests...", files: ["templates/k8s-deploy.yaml"] }

Notes

Skills are per-session (uploaded to session filesystem)
Path traversal protection on skill names
Invalid SKILL.md files are reported but don’t block discovery of other skills
For organization-wide skills, see the Skills Registry