Skip to main content

Documentation Index

Fetch the complete documentation index at: https://www.truefoundry.com/llms.txt

Use this file to discover all available pages before exploring further.

Pick the skills you want, save the agent, and the Gateway handles the rest at runtime. Skills require an agent with the Sandbox enabled.

​
Mount Skills in the Playground

The Agent Playground exposes skills as a sidebar section, next to MCP Servers.
1

Open the Playground for a TrueFoundry Agent

Navigate to Agents → Playground and either create a new agent or open an existing one. Confirm the Sandbox toggle is enabled.
2

Open the Skills selector

Click the + button next to Skills in the left sidebar. The selector lists every skill you have access to, grouped by Repository. Each row has a version dropdown — defaults to the latest version, but you can pin any older one.
Skills selector in the Agent Playground listing available skills with version pickers and a preload toggle
3

Pick the skills the agent needs

Check the skills you want. Each selected skill becomes a skills[] entry on the agent.
Within a single agent, two skills cannot share the same name even if they live in different Repositories. The selector greys out the second skill and shows a tooltip explaining the conflict — pick one or use a differently named copy.
4

(Optional) Toggle Preload SKILL.md per skill

Each selected skill has a small icon toggle that controls Preload SKILL.md:
ModeBehavior
Off (default)Only the skill’s name and description are exposed upfront. The body is loaded on demand when the agent decides the skill is relevant.
OnThe full SKILL.md body is loaded upfront. Higher upfront context cost, but no extra step at runtime.
Turn preload on for short, always-relevant skills (e.g. a 200-line style guide). Leave it off for long, situational skills that only fire occasionally.
5

Test, then save the agent

Send a prompt that should trigger the skill, then click Save Agent to persist the skill list. Subsequent runs of the agent mount the same skills automatically.

​
Mount Skills via the Agent API

If you’re invoking agents programmatically, pass the skills in the request body. The shape is identical for saved agents and inline agents.

​
Request Shape

{
  "skills": [
    {
      "fqn": "agent-skill:my-tenant/my-skills/analytics-helper:1",
      "preload_skill_md": false
    },
    {
      "fqn": "agent-skill:my-tenant/my-skills/hello:2",
      "preload_skill_md": true
    }
  ]
}
FieldTypeRequiredDescription
skills[].fqnstringyesSkill version FQN in the form agent-skill:<tenant>/<ml_repo>/<name>:<version>. The version segment must be a positive integer — :latest is not supported.
skills[].preload_skill_mdbooleanno, default falseWhen true, the full SKILL.md body is loaded upfront. When false, only name and description are exposed; the body is loaded on demand.

​
Constraints

  • Sandbox required. Inline agent requests must set sandbox.enabled = true. The Gateway rejects requests with skills but no sandbox.
  • Unique names per request. Two skills in the same request cannot share the same name, even across repositories. The agent rejects the request before invoking the model.
  • RBAC. The caller must have read access on each skill’s Repository. For requests that target a saved agent (via agent_name), the Gateway treats READ on the agent as implicit access to the skills attached at save time.