research-reportlisted
Install: claude install-skill kennykankush/skillpack
# research-report
Two-mode research workflow. Same engine, two output destinations.
## When to use
Trigger on:
- `$workbench:research-report official <topic>` — Codex official mode (files written, persistent)
- `$workbench:research-report scan <topic>` — Codex scan mode (chat reply only, ephemeral)
- `/workbench:research <topic>` — official mode (files written, persistent)
- `/workbench:scan <topic>` — scan mode (chat reply only, ephemeral)
- Natural language: "research X", "deep dive on Y", "give me a quick read on Z", "scan W", "what's the state of V", "write up findings on U"
Skip when:
- Quick fact lookup → use WebSearch directly
- Single-link extraction → use WebFetch directly
- Code question about a known library → use docs lookup
- Anything where the answer is one paragraph → just answer inline
## The two modes
### Official mode (`/workbench:research`)
Full pipeline: research, write notes, synthesize report, render HTML, open in browser.
End state: exactly two files in `research/<umbrella>/<title>/`:
- `notes.md` — consolidated raw research dump (one file even if many sources covered)
- `report.html` — polished, designed, render-ready
Hidden in `.build/` subdirectory: `report.qmd` source, quarto cache. Never visible to user.
### Scan mode (`/workbench:scan`)
No files. Replies in chat with structured markdown:
```markdown
## Key findings
- Bullet 1 (the punchline, not the setup)
- Bullet 2
- Bullet 3
## Sources
- [Link with one-liner of why it's relevant]
-