ai-prospecting

Wraps ai_prospecting — the “who are the best people to approach at this account” tool — and the dataset work that follows a run.

When the agent picks this skill

• The user wants buyers, champions, or a “who should I talk to” list at one or many companies, and wants the engine to decide who.

Not for deterministic lookups by title, persona, seniority, location, or technology — those route to ask_onfire over the contact entity. There is no single-person scoring path.

What the skill teaches the agent

• The single action (action="run"), single-company and batch forms, and the shadow-schema test mode.
• The polling pattern: a still_running response carries run_ids — pass them back rather than re-submitting.
• How to read top_picks, priority_summary, and facets, and how to query the run’s dataset for anything deeper.
• Hand the run’s dataset_id straight to contact_data_enrichment instead of rebuilding contact rows.

Underlying MCP tools

ai_prospecting · ai_prospecting_field_glossary · query_datasets · describe_dataset · download_dataset