helpin-ai/helpin/helpin-delegate-agent

Delegate to Helpin

1. Call get_current_context and list_agents.
2. Match the requested outcome to an agent's role and supported targets. Ask the user when several agents or targets are plausible.
3. Confirm the exact agent_id, target_type, and target_id. Do not invent IDs or select a different workspace.
4. Call start_agent_run once with a stable idempotency key and only the context the agent needs.
5. Poll get_agent_run; do not start duplicate runs while status is queued, running, or paused. Explain that v1 uses polling and has no completion webhook.
6. If the run needs approval or input, stop and present that state to the user. Cancel only on an explicit user request.
7. Finish with run status, artifacts, output summary, and Helpin links returned by the tools.

If agent-run scope is unavailable, do not imitate a successful delegation. State which capability is missing and offer to perform only the safe local analysis.