Track topics with a personal intelligence agent
A user turned Claude Code into a personal intelligence agent using CLAUDE.md and skills for data collection and analysis, watching topics and surfacing things they might miss. Task: Suggested implementation: Build a monitoring agent in Claude Code using CLAUDE.md for behavior and separate skills for data collection and analysis. The agent tracks topics defined in plain language, replacing manual checks of sources like HN, GitHub Trending, and arXiv. The user provides feedback to prioritize findings and reviews generated reports.; tools: Claude Code. Human role: The original post does not say how a person stayed involved.; limitations: Source is inaccessible (HTTP 403), so all details are unverified. Prerequisites include configuring CLAUDE.md and developing sensor/lens skills. Integration uncertainty exists regarding API access for external sources. Human approval is required for final report distribution..
Tools
Published on Agent Work Atlas:
Overview
A user turned Claude Code into a personal intelligence agent using CLAUDE.md and skills for data collection and analysis, watching topics and surfacing things they might miss.
Workflow
Suggested implementation: Build a monitoring agent in Claude Code using CLAUDE.md for behavior and separate skills for data collection and analysis. The agent tracks topics defined in plain language, replacing manual checks of sources like HN, GitHub Trending, and arXiv. The user provides feedback to prioritize findings and reviews generated reports.
Reported result
Outcome unknown due to inaccessible source. Proposed acceptance check: Measure time reduction from manual scanning to report review, targeting a shift from ~1 hour to ~2 minutes per day. Verify report relevance against manual scans for accuracy.
Human role
The original post does not say how a person stayed involved.
Limitations
Source is inaccessible (HTTP 403), so all details are unverified. Prerequisites include configuring CLAUDE.md and developing sensor/lens skills. Integration uncertainty exists regarding API access for external sources. Human approval is required for final report distribution.
Try it
Act as a technical writer creating a setup guide for a personal intelligence agent in Claude Code. Context: Users want to automate monitoring of tech topics (HN, GitHub, arXiv) using plain language definitions. Target: Developers familiar with Claude Code. Output: A step-by-step guide in Markdown. Steps: 1. Define the agent's role in CLAUDE.md. 2. Create a 'sensor' skill to fetch data from specified sources. 3. Create a 'lens' skill to analyze and summarize the data. 4. Explain how to provide feedback to the agent. 5. Include a dry-run checklist to verify output against manual scans. Deliverables: CLAUDE.md snippet, skill structure example, and feedback loop instructions. Success metric: The guide must enable a user to set up the agent in under 30 minutes. Handle missing inputs by assuming standard API access for the mentioned sources. Do not claim this was tested; label it as a proposed workflow.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.