deep-researchlisted

# Deep Research Produces a **single `dashboard.html`** per run, saved to a timestamped directory. Chat output is always a short summary block pointing to it. --- ## Setup (before anything else) 1. Determine the output directory: - If inside a project: `[project root]/research/YYYY-MM-DD-HH-MM-[query-slug]/` - Otherwise: `~/research/YYYY-MM-DD-HH-MM-[query-slug]/` - `query-slug` = first 4-5 words of query, lowercased, hyphenated 2. Create the directory with `mkdir -p` 3. State the full output path before proceeding Deliverables are decided in Phase 1 based on query complexity and type. The only constant: keep chat output to a short summary block. Never paste full research into chat. --- ## Phase 1: Query Analysis Classify the query. Write this analysis in chat (briefly) before proceeding. **Type:** - `comparative`: X vs Y vs Z - `exploratory`: state of X, how X works - `temporal`: history/evolution of X - `decision`: should we do X - `factual`: specific facts, numbers - `synthesis`: best approaches to X **Answer shape:** what does a correct answer look like as a data structure? (table / timeline / ranked list / brief+evidence / narrative) **Failure mode:** what would make this answer wrong or useless? (outdated sources / biased sources / missing key dimension) **Answer format:** what is the most natural structure for the answer? (narrative, ranked list, recommendation+evidence, timeline, comparison) **Deliverables:** decide here what files to produce, bas