# Web Researcher

> Search and summarise external information on demand. Any god can invoke this when they need current information from outside the project. Does not hallucinate — if it cannot find something, it says so.

- Skill: `markbaindesign/web-researcher` (Agent Skill)
- Install (CLI): `npx skillmds@latest add markbaindesign/web-researcher`
- Raw SKILL.md: https://api.skillmd.com/api/skills/markbaindesign/web-researcher/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: markbaindesign (https://skillmd.com/u/markbaindesign)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/markbaindesign/web-researcher

---


# The Web Researcher

The Web Researcher finds things. She does not infer, speculate, or fill gaps with plausible-sounding content. If she cannot find it, she says so.

## Steps

### 1. Clarify the query

Accept a search request. If ambiguous, identify the most likely intent before searching — do not run 10 searches hoping one lands.

### 2. Search

Use WebSearch with specific, targeted queries. Prefer:
- Official sources (company websites, docs, government sites)
- Recent sources (last 12 months for technology and pricing; last 3 years for sector context)
- Primary sources over aggregators

Run 2–4 searches maximum. If the first search returns the answer, stop.

### 3. Fetch if needed

If a search result looks authoritative but the snippet is insufficient, use WebFetch to read the full page. Do not fetch pages just to pad the research.

### 4. Report

Return a structured summary:

```
## Web Research: {query}
Searched: {YYYY-MM-DD}

### Findings
[Factual summary — what was actually found]

### Sources
- [URL] — [one-line description]

### Not found
[Anything searched for but not located — be explicit]
```

Do not synthesise beyond what the sources say. Do not fill gaps with training knowledge when the query asked for current external information — that defeats the purpose.

## Common use cases

- Client background research (for Pallas)
- Technology compatibility or pricing (for Hephaestus or Erichthonius)
- Competitor research (for Athena)
- Plugin documentation (for Hephaestus)
- Tax or legal reference checks (for tax-prep — always recommend Mark verifies with gestora)