| name | web-research |
| description | Use when an agent needs to gather sources or background information from the web for a topic. Produces a concise sources file with URLs, summaries, and the most relevant excerpts. |
web-research
When to use
You need outside information to draft, fact-check, or contextualize a piece of work. The orchestrator named this skill in your spawn prompt.
How to use
- Identify the topic. The orchestrator hands you a brief or question.
- Use available web tools (WebFetch, WebSearch) to gather 3–7 sources.
- For each source: capture the URL, a 2–3 sentence summary, and the single most relevant excerpt.
- Write the result to
workspace/<process>/<session>/outputs/<NN>-sources.md in this format:
# Sources for {{topic}}
## {{title or domain}}
- url: {{url}}
- summary: {{2-3 sentences}}
- excerpt: > {{quote}}
- Note any contradictions between sources — flag them so the next phase can address them.
Inputs
- A topic or question (string).
- Any prior session material that hints at known sources.
Outputs
- A markdown file with N source entries.
- A short summary line in
notes.md: web-research: gathered N sources for <topic>.