Instead of generating candidate lists and leaving the rest to humans, it: * converts a business goal into explicit role specs (deliverables, success metrics, disqualifiers) * prioritizes agent-based solutions before sourcing humans * enriches each candidate with structured, evidence-backed data * scores candidates using a rubric with hard caps on missing requirements * enforces pipeline transitions through a PM agent with guarded state changes
All state is persisted in SQLite. All workflows are defined in YAML. No hidden context, no implicit transitions.
The pipeline behaves like a state machine:
new → researching → ready_to_contact → screening → trial → hired / rejected
The broader question I’m exploring is whether staffing should be treated more like infrastructure — with explicit state, constraints, and evaluation gates — instead of a sequence of loosely connected conversations.
Interested in feedback on:
* rubric design and failure modes * agent-first vs human-first sourcing * where deterministic workflows break down in hiring