intake

Pipeline Intake — rough idea to triage-ready briefs

Turn a rough idea into one-or-many briefs the triage stage can pick up. A brief is small on purpose: enough for triage to ground and judge, never a spec. This stage is where the pipeline is generous — divergent, exploratory, audience-minded — because everything after it narrows.

Running as a Gemini model? Read gemini.md in this directory first, then follow this file with the overrides it names. It gives step 4's 'features the audience would likely want' a counted ledger so the divergent pass cannot land as one brief, turns 'briefs, not specs' into a grep run over the written files, makes trawl and the research panel emit artifacts step 5 has to open, and requires every research claim to carry its citation check and its cost. Other models skip it.

Inputs

• A rough idea: inline text, a file, a voice-note transcript, a routed directive. Anything from "an app that does X" to "we should probably handle Y".
• Optional: a target repo (default: the current one). The briefs land in that repo's docs/features-to-triage/.

Procedure

1. Read what exists. docs/features-to-triage/, the ledger/board, and any docs/deep-research/ reports — so a new brief extends the backlog instead of duplicating it. Match by concept, not wording ("night theme" duplicates dark-mode). If the repo keeps an out-of-scope record (.out-of-scope/ or similar), check it: a previously rejected concept is surfaced to the user ("this was rejected before because — still true?"), not silently re-added.

2. Research where the idea is thin. A rough idea usually leaves open questions about the world — market norms, prior art, what competing products do, what the audience already uses. Those are research questions, not guesses: run them per ${CLAUDE_PLUGIN_ROOT}/references/second-opinion-lanes.md §Deep research (Dossier research_plan → panel, free lane first; verify citations; export to docs/deep-research/<slug>.md). Skip research when the idea is already concrete and internal.