start-research

# Start-Research Interview-style flow that takes a freshly-installed OMCR project (post-`/omcr-setup`) and walks the user through filling in **what their research is about** — working title, hypothesis, target venue, datasets, narrative spine, preset choice — then scaffolds the LaTeX manuscript directory. **When this skill is invoked, immediately execute the workflow below. Do not only restate or summarize these instructions back to the user.** This is the "deep interview" complement to `/omcr-setup`'s "install" role. ## Argument parsing Inspect `$ARGUMENTS` (passed through by [`commands/start-research.md`](../../commands/start-research.md)). Treat the **first token** as the preset hint: - `minimal` — skip the preset-overlay prompt entirely (no `examples/<field>/` overlay) - `neuro-fmri` — pre-select the neuro-fMRI preset overlay (still confirms before applying in phase 2) - `no-overlay` — alias of `minimal` - (empty) — ask the user during the interview (phase 2) Additional standalone flags (may appear in any position; not mutually exclusive with the preset hint): - `--no-venue-seed` — skip phase 5 (venue scope seed). `@reviewer` stays generic. Record this so phase 7 surfaces it. Record the preset hint as `preset_arg` and any flags as `flags` so downstream phases can consume them. ## Phase execution Execute phases sequentially. For each phase, read the linked file and follow its instructions exactly. If an earlier phase reports an early stop (e.g. user cancelled, `/omcr-setup` not run and user declined to run it), do **not** skip phase 7 (the report) — surface the stop reason there. 1. **Phase 1 — Precheck.** Read [`phases/01-precheck.md`](phases/01-precheck.md). 2. **Phase 2 — Interview.** Read [`phases/02-interview.md`](phases/02-interview.md). 3. **Phase 3 — Fill CLAUDE.md.** Read [`phases/03-fill-claude-md.md`](phases/03-fill-claude-md.md). 4. **Phase 4 — Preset overlay (agent memory).** Read [`phases/04-preset-overlay.md`](phases/04-preset-overlay.md). 5. **Phase 5 — Venue scope seed (reviewer specialization).** Read [`phases/05-venue-scope-seed.md`](phases/05-venue-scope-seed.md). Runs **after** phase 4 on purpose: if phase 4 installed a preset reviewer `MEMORY.md`, phase 5 appends the venue block to it without disturbing the preset content. If phase 4 didn't run (no preset), the canonical template is still byte-identical to the schema, so phase 5 performs a full structured replacement instead. 6. **Phase 6 — Manuscript scaffold (delegated).** Read [`phases/06-manuscript-scaffold.md`](phases/06-manuscript-scaffold.md). This phase delegates to the `manuscript-scaffold` skill. 7. **Phase 7 — Report.** Read [`phases/07-report.md`](phases/07-report.md). ## Re-running policy Safe to re-run on an initialized project. The re-run contract: - Phase 2 surfaces existing `CLAUDE.md` values as defaults. The user can confirm-all to apply only newly-added fields. - Phase 3 writes only fields that changed. - Phase 4 only replaces agent `MEMORY.md` files that are still byte-identical to the canonical template (untouched). Modified memory is never overwritten. - Phase 5 leaves an existing `## Venue-specific bar` section in `reviewer/MEMORY.md` alone. To refresh, delete just that section and re-run. - Phase 6 delegates to `manuscript-scaffold`, which has its own existing-content guard. If the user wants to **reset** an agent's memory, instruct them to delete that specific `.claude/agent-memory/<agent>/MEMORY.md` and re-run `/start-research`. There is intentionally no `--reset` flag — deletion is an explicit, reviewable action. ## Final recommendation After phase 7's report, end by recommending the user run `@supervisor where are we?` for the first real conversation.