Starter prompt execution reference

Execution reference: This page specifies the tool protocol for automation agents. For instructions to follow in Studio, use the user guide.

Studio starter prompts route project tasks through the veryfront skill. Agent reads the matching reference, then uses live project and catalog tools.

Examples show project source or a saved artifact first, followed by the MCP call when applicable.

Studio promptCanonical docsExpected first actions
Create a planCreate a planLoad veryfront, read the docs, ask one brief field when details are missing
Create an agentCreate an agentLoad veryfront, read the docs, search integrations, create or update the agent
Create a skillCreate a skillLoad veryfront, read the docs, create or update a reusable skill

External web search and web fetch are unavailable in Veryfront Agent. For a direct research request, use supplied sources and authorized project documentation.

Runtime contract

  • Load veryfront__load_skill({ "skillId": "veryfront" }) first.
  • Do not load removed sub-skills such as create-agent, plan, research, create-skill, or create-agentic-workflow.
  • Read only the canonical docs page needed for the active prompt.
  • Prefer live tool results for project state, integration IDs, tool IDs, files, agents, prompts, resources, and workflows.
  • Ask exactly one required freeform brief field for broad first-turn intake when details are missing.
  • Do not create app, page, route, or chat UI files unless the user asks for those artifacts.

Verification

Verify each current starter in a clean conversation through a Veryfront E2E test or staging run. Check for visible output and no terminal error. Retain the evidence with the issue before closing it.