researchlisted

# /research - Project & Technical Investigation Investigate the codebase and record findings with sources, without implementation. ## Input The research subject is taken from `$ARGUMENTS`, a free-text topic or question. If empty, prompt via AskUserQuestion. ## Source notation Facts are `file:line` or command output, inferences `inferred from X`, unverified `unknown, requires X`. This is the source notation the Phases and the report template refer to; no other form is accepted. ## Phase 1: Outcome Anchor Read `.claude/OUTCOME.md`. If absent, generate the stub via /outcome. If the investigation steps into Non-goals, confirm with the user before proceeding. ## Phase 2: Prior Research Scan Derive the lowercase hyphenated subject slug from `$ARGUMENTS` and run `${CLAUDE_SKILL_DIR}/scripts/find-prior-research.py <slug> .claude/workspace/research`. Parse the JSON `{ candidates: [{file, shared}, ...], slug_words: int }` (shared descending) from stdout. - No candidates: set the report's Prior research to `none found` and move on - A candidate with shared >= 2, or with shared equal to `slug_words`: carry forward per the table below - A candidate with shared == 1 left after that: the slug runs two words or more, so the filename overlap is all it rests on. It is excluded from the carry-over table below and lands only in the report's References, with its path and shared count | Extract | Carry to | Handling