P1 Spec Research

• Phase 0 (goal-clarifier): when $ARGUMENTS is a sparse seed, the orchestrator first dispatches goal-clarifier for an ambiguity-scored interview across 4 RTL dimensions (Functionality / PPA Target / Scope / Verification). Phase 0 writes docs/phase-1-research/goal.md.
• Phase 1 (spec-analyst + research): spec-analyst consumes goal.md (plus any user-supplied spec document) and produces the iron/open requirements set as before.

When $ARGUMENTS points to an existing .md/.txt/.rst spec or is already a rich seed (≥ 500 chars with PPA/coverage signals), Phase 0 is skipped automatically.

Prerequisites

None — this is the first phase entry point.

Execution

Task(subagent_type="rtl-agent-team:p1-research-orchestrator", prompt="Execute Phase 1 spec research. User input: $ARGUMENTS")

Do not perform any work directly. The orchestrator agent manages spec refinement, solution tree exploration, sub-domain expert coordination, and 3-round chief review.

Output Artifacts

• docs/phase-1-research/iron-requirements.json — settled functional/performance requirements (REQ-F-, REQ-P-)
• docs/phase-1-research/open-requirements.json — research topics deferred to Phase 2 (OPEN-1-*)
• docs/phase-1-research/io_definition.json — I/O interface definitions
• Classification verification step confirms each requirement is tagged as either iron (acceptance_criteria defined) or open (research_needed).