Runtime vs. Preview vs. Evaluation

Prompt Orchestra distinguishes between runtime, preview, and evaluation surfaces in the public platform.

Runtime

• POST /api/skills/:id/run -> public beta
• POST /api/agents/:id/run plus GET /api/agent-runs/:id -> stable

Evaluation

• POST /api/skills/:id/evaluations/run -> public beta

Preview

• POST /api/skills/:id/preview -> public beta

Use preview when you want composition truth without model execution or state mutation.

Public beta skill docs now treat all three loops as supported:

• GET /api/skills/:id
• POST /api/skills/:id/preview
• POST /api/skills/:id/run
• evaluation routes

See also

• Run A Skill
• Run An Evaluation
• Public Boundary