What problem does it solve? Raw ideas and brain dumps arrive ungrounded, and designing on them directly risks missing existing code, prior team decisions, or outside prior art. This Skill turns a raw idea into a structured exploration document before anyone commits to a design. ## Core Features & Use Cases - Parallel fan-out dispatch: Sizes and dispatches 3-12 subagent workers across three legs — internal investigation, external research, and prior-decision recall from the ZZ knowledge base — one question per worker. - Synthesis into explore.md: Writes a single initiative document with Background, Current state, and Rough direction sections, including 3-5 ranked candidate directions spanning at least two resolution shapes, each with citations or explicit sentinels. - Degraded-mode handling: Defines sentinel values and failure rules for each worker leg, including when not to write the artifact at all. - Use Case: A product manager brings a vague feature request about token expiry. The Skill captures the brain dump, dispatches investigations into the auth module and session layer plus a recall of prior decisions, then synthesizes ranked directions for the team to pick from before writing a spec. ## Quick Start Ask the agent to run sdlc-explore on your raw feature idea so it can fan out investigators and produce an explore.md with ranked directions.
