Content Researcher
Deep research subagent for Bryan Collins' content operation — gathers sources, frameworks, statistics and competitive analysis for articles, newsletters and YouTube scripts. Research only, no writing.
Entry verified April 21, 2026
The short answer
Runs with Read, Glob, Grep, WebSearch and WebFetch — no write access. Handles credible stats, competitive analysis, framework research (Porter's, PESTLE), source vetting, and keyword intent. Returns a structured brief with core angle, key facts, coverage gaps, an H2 outline and source URLs.
When to use it
Use before substantial content pieces where the writer needs a vetted source pack. Useful when the topic risks hallucinated citations, or when running research in parallel with drafting.
Setup
- 1
Save the file as ~/.claude/agents/content-researcher.md.
- 2
Frontmatter grants Read, Glob, Grep, WebSearch and WebFetch — deliberately no write tools.
- 3
Output is a five-part brief: core angle, key facts with sources and dates, gaps in coverage, suggested H2 outline, full URLs with access dates.
- 4
Tuned for Bryan's channels — newsletter (Tue/Thu), LinkedIn (Mon/Wed/Fri), YouTube — and the "content is exhaust not engine" philosophy.
- 5
Cite the actual study, not "studies show" — the agent is instructed to reject vague attribution.
Example
You: Research 'Claude Code for writers' Researcher: [fetches 12 sources, returns 600-word brief with 8 cited stats, 3 competitor angles, and 1 contrarian take]
Source & attribution
- Author
- Bryan Collins (portfolio pattern)
- Licence
- MIT
- Type
- Original
Original pattern published under MIT — attribution preserved by convention, not licence requirement.
Caveats
Web fetch can fail on paywalled sources. Configure fallback behaviour explicitly.
56 skills across 10 categories, all licence-verified.