Content Research Pipeline
From keyword to published article — orchestrate research agents, fact-checkers, and image generation in a single Chro workspace. No copy-paste between tools.
Content teams, knowledge workers
1. Primary Research
Run a Codex agent with the target keyword. It collects definitions, mechanisms, related concepts, and a list of primary source URLs — all in one pass.
Investigate
[target_keyword]: what it means, why it matters, how it works, related keywords, and primary sources.
Output lands in the inbox as structured Markdown. No manual search needed.
2. Source Collection
Launch parallel worktree agents to fetch official sources — papers, docs, GitHub repos — identified in step 1. Each agent scrapes one URL and outputs clean Markdown.
- Agent A: arXiv paper → extracted abstract, methodology, results
- Agent B: official documentation → key API surface, architecture
- Agent C: GitHub README → project scope, setup, benchmarks
All three run simultaneously. Raw files merge into a single research folder.
3. Organize & Draft
A summarizer skill reads the entire research folder and produces a structured draft following your internal format: overview, technical deep-dive, use cases, and key takeaways.
The draft auto-links to source files so every claim is traceable.
4. Fact-Check & Review
Run the fact-check skill against the draft. It cross-references claims with the collected primary sources and flags discrepancies.
- Verified: 23 claims
- Needs source: 2 claims
- Corrected: 1 date, 1 metric
A second pass with the finalizer skill polishes tone, fixes formatting, and produces the publication-ready version.
5. Thumbnail Generation
The slide skill extracts a 3-column visual summary from the article. An image agent sends the prompt to Nano Banana Pro at 2K resolution and embeds the result back into the document.
Keyword → research → draft → review → publish — one workspace, zero context-switching.