Claude Code researches attention with citation checks
Keiffenheim reports using her Jean Claude Code skill to produce a report on phones, social media and attention, with scoped searches and citation checks before using it for further reading.
evakeiffenheim.substack.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Research [attention question] within [academic domains]. Scope depth first, check paper identities and DOI/URL links, preserve competing explanations and open questions, and produce a report for further source reading.
Reach people exploring AI agents
See placements →What happened
Keiffenheim reports using her Jean Claude Code skill to produce a report on phones, social media and attention, with scoped searches and citation checks before using it for further reading.
Reported by Eva KeiffenheimHow the agent did it
- Scope the attention question, academic domain and desired depth.
- Search databases, then relevant authors, journals and citation trails.
- Check bibliographic identity before retaining each citation.
Where a person stepped in
Keiffenheim uses the report as a starting point for further reading. The full skill implementation is not inspected here, and no independent citation-accuracy audit is supplied.
Eva Keiffenheim adapts an academic-research skill for Claude Code and names it Jean. She reports using it to investigate what phones, social media and notifications have done to attention, then links the resulting research PDF and a post it informed.
The workflow first scopes the domain, desired depth and focus. It searches academic databases, expands toward authors and journals, and follows citation trails. It assesses relevance, authority, recency and source type, and requires each cited paper's title, author, year and DOI or URL to be checked before the report is written. Unverified citations are removed under the stated protocol.
Keiffenheim receives the attention report after roughly twenty minutes and uses it as a starting point for further reading. That is a concrete research artifact, not independent proof that all conclusions or citations are correct. The public portion of the post describes the method and outcomes; the full downloadable skill and installation guide sit beyond the accessible section, so this entry does not claim to inspect their implementation. The useful boundary is to scope before search, preserve debates and open questions, and treat the generated report as a guide back to sources rather than a final authority.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Blog · Eva Keiffenheim
Frequently asked questions
Which completed report is identified?
The author links a report on phones, social media, notifications and attention, and a subsequent post it informed.
What happens to an unverified citation?
The stated protocol removes it unless its title, author, year and actual DOI or URL can be confirmed.
Was the full skill inspected?
No. The publicly accessible section explains the workflow, but the full installation and skill materials are beyond that section.
More details
Remaining steps
- Produce the report with debates and open questions visible.
- Use the linked PDF to guide further human reading.
About this account
Claude Code follows the Jean skill to scope, retrieve, evaluate and cross-check academic sources before preparing a research report.
Firsthand account by Eva Keiffenheim; original publication date verified against Primary Substack May 18, 2026. Outcomes are the author’s reports, not independently reproduced. Suggested prompt written by Agents In Use; not the author’s original prompt.
Credit: Eva Keiffenheim
Source published 18 May 2026
Source checked 14 Oct 2026
Tools and topics
Claude Code · Jean research skill · Semantic Scholar · arXiv · Research PDF
Attention research · Citation identity · Further-reading scaffold
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.