deep-research-planlisted

# Deep Research Plan Research first; planning and execution still follow `../optim-plans/SKILL.md`. ## Required Flow 1. Read and follow `../optim-plans/SKILL.md`, treating the request as if it named `deep-research-plan`. 2. Inspect the target Git repo read-only and initialize or resume the optim-plans controller before external research. 3. Ensure the root `.gitignore` ignores `refs/`; add that ignore entry only during execution approval if it is missing. 4. Search proactively for related projects, articles, and papers. Prefer `agent-reach` when available. If `agent-reach` is missing, ask exactly one yes/no sentence about installing it locally; on yes, install it first and use it, and on no, use websearch fallback and record the fallback reason. 5. Download at least 3 relevant refs into `./refs/<topic>/` before writing `PLAN_v1.md`. Do not satisfy this with `curl` alone, README-only snapshots, or abstract-only paper notes; keep full local project, article, paper, docs, or artifact content sufficient for deep analysis. 6. Record every downloaded ref through the controller with `deep-record-ref` before relying on it. Record graph output with `deep-record-graph`, analysis artifacts with `deep-record-analysis`, and graphify install refusals with `deep-waiver-question` plus `deep-record-waiver`. 7. For every downloaded ref, generate graphify JSON beside the ref before user adoption questions. If `graphify` is missing, ask exactly one yes/no sentence about installing it locally;