research

Execute each section in order. Copy the checklist and check off items as you complete them. Do not proceed past a Gate until verified.

Artifact: Write to .work/research/. Verify with ls before wrapping up.

Constraint: You MUST NOT modify project files. This skill produces context, not code.

Mode

Check .work/active/ for marker files. If exactly one exists and no $ARGUMENTS provided, auto-select that feature's slug. If multiple markers exist, list them and ask. Arguments always override.

Parse $ARGUMENTS (or auto-selected slug): matches existing file in .work/research/ by slug → re-entry. Set but no match → create. Empty → ask what to research, then proceed in create mode.

Brainstorm required: Find matching brainstorm artifact in .work/brainstorms/. If missing, print "Run /dl:brainstorm first." and stop. Read the brainstorm artifact and extract the ## Research Queries section.

Task Progress:
- [ ] Determine mode; find and read brainstorm artifact
- [ ] If brainstorm missing: stop with message
- [ ] If re-entry: read existing research artifact
- [ ] Run /resolve-rules mode:keyword <topic>
- [ ] Execute each research query as targeted search
- [ ] List findings per query in response
- [ ] Synthesize: identify gaps across queries
- [ ] Write artifact (create: new file; re-entry: append dated section)
- [ ] Gate: verify artifact with ls; update active marker
- [ ] Wrap up: summarize, suggest /dl:plan

Query execution

Run /resolve-rules mode:keyword <topic> as a subtask. For each query from the brainstorm artifact, search the codebase with targeted reads and greps. For linked repos from resolve-rules, include them in the search. Record findings per query — keep findings factual, not opinionated. When a query touches an existing feature, note how it's structured through the stack (e.g., route → service → model → test) — this gives plan concrete examples for vertical slicing.

Synthesis

Compare findings across queries. Identify gaps — things the brainstorm assumed that don't match reality, missing implementations, inconsistencies. These become the Gaps & Recommendations.

Re-entry

Read the existing research artifact in full. Execute brainstorm queries again (codebase may have changed since last session). Append a new ## YYYY-MM-DD dated section — do not overwrite prior sections.

Output

Write to .work/research/YYYY-MM-DD-<slug>-research.md:

# <Topic> Research

## YYYY-MM-DD — <description>

### Applicable Rules
| Rule | Source | Key Patterns |

### Query: <research question>
- **Finding:** <what was discovered>
- **Location:** <where>
- **Implications:** <what this means>

### Gaps & Recommendations
- [ ] <actionable item>

Repeat ### Query: for each research query. Omit ### Applicable Rules on re-entry if rules haven't changed.

Gate

Verify artifact with ls; write if missing. Update .work/active/<slug>.md with stage: research and today's date.

Wrap up

Summarize: rules matched, findings per query, gaps count. Suggest next step: /dl:plan <slug>.

Rules

• Do not modify project files. This skill produces context, not code.
• Do not overwrite prior research sections on re-entry (findings accumulate).
• Keep findings factual — opinions belong in brainstorm decisions.
• If no brainstorm artifact exists, stop and tell the user.