Skip to content
OnFire OnFire Docs

manage-ai-prospecting

Edit the caller’s prospecting schema via the shadow lifecycle. Iterate in chat → test with ai_prospecting(activate_shadow_run=True) → promote when satisfied. All operations are scoped to the caller from OAuth — no tenant / email arguments.

When the agent picks this skill

Section titled “When the agent picks this skill”
  • “Show me my prospecting schema / regions / keywords / ranking config.”
  • “Tweak hot keywords and let me try the result before going live.”
  • “Promote my draft to live.” / “Throw away the draft.”

Not for triggering prospecting runs (use ai-prospecting) or inspecting run history.

What the skill teaches the agent

Section titled “What the skill teaches the agent”
  • The full lifecycle: get_my_schema → create_shadow → update_shadow (× N) → test → ship_shadow(confirm=True) (or discard_shadow).
  • Always start with get_my_schema to read the live config before proposing edits.
  • The editable fields under patch and which ones are protected server-side.
  • Never call ship_shadow without both an explicit user approval phrase (“ship it”, “promote it”, “make it live”) and confirm=True.
  • The shortcut-row caveat: create_shadow on a solo-team schema flips is_shadow=TRUE in place; discard_shadow flips it back but any edits made during testing remain on the live row.

Underlying MCP tools

Section titled “Underlying MCP tools”