sdkHosted AI, explicit consent

Run skill

pipeline.runSkill

Run a Pipeline-owned first-party skill through Pipeline hosted inference. This consumes the workspace BYOK model provider, not the coding agent subscription. For execution inside Claude Code, Codex, Hermes, or another external agent, claim an Agent Teams work order instead. Returns a hosted run summary including the runId for get-agent-run.

Parameters

workspaceId
string
Optional workspace ID or slug. Required when this credential can access multiple workspaces and no workspace was configured on the client.
skillNamerequired
string
The skill name from frontmatter (e.g. dream-one-hundred, cold-email-write).
payload
object
Arbitrary input passed to the skill.
agentRole
string
Optional: pin which agent role context to use.
scout | researcher | icp | strategist | writer | campaign_builder | closer | coach
allowHostedAi
boolean
Must be true when this operation uses Pipeline hosted inference.

Example

TypeScripttypescript
await pipeline.runSkill({
"skillName": "example",
"allowHostedAi": true
})

Response

Success response schemajson schema
{
"type": "object",
"properties": {
"id": {
"type": "string"
},
"status": {
"type": "string"
},
"terminationReason": {
"type": "string"
},
"toolCalls": {
"type": "number"
},
"tokensUsed": {
"type": "number"
},
"skipped": {
"type": "boolean"
},
"reason": {
"type": "string"
},
"roleDispatched": {
"type": "string"
}
},
"required": [
"id",
"status",
"terminationReason",
"toolCalls",
"tokensUsed"
]
}

Exceptions

PipelineError
Thrown for validation, authentication, JSON-RPC, HTTP, and network failures. Inspect code and data when present.
Pipeline TypeScript SDK documentation | Pipeline