What this agent does
Brand mentions matter most in the first few hours — a public complaint that gets a fast reply reads very differently from one that sits ignored for three days — but nobody has time to search their brand name manually every morning. This agent watches for new mentions and separates the ones that need same-day attention from the ones that can wait for your weekly digest.
mention-searcher— searches X/Twitter and Reddit for new mentions of your brand, product name, and configured aliases since the last run.urgency-classifier— scores each mention as urgent (public complaint, bug report, or a post gaining rapid traction), notable (relevant discussion, competitor comparison), or routine (passing mention), and sends an immediate alert for anything urgent while batching the rest into a daily digest.
Every run appends to a mentions log with the classification and a link to the original post, so nothing found gets lost even if it wasn't urgent enough for an immediate alert.
Requirements
- Runs on free, built-in search tools (
x_search,web_search/Reddit search) — no paid social-listening API required. - Works on Claude AI (claude.ai or Claude Code) and on Zo Computer natively.
Setup
- Unzip the package into your workspace, keeping the folder structure intact.
- Edit
config/monitor-config.mdwith your brand name, product aliases, and the channel where urgent alerts should be sent. - Run the starter prompt in
starter-prompts.mdto do a first pass and confirm the aliases catch real mentions without too much noise. - Schedule
automation-prompt.mdto run every 1-2 hours for the urgency check, with a separate daily run for the full digest.
What's included
Skills/mention-searcher/SKILL.mdandSkills/urgency-classifier/SKILL.mdconfig/monitor-config.md— brand name, aliases, and alert channelconfig/mentions-log.md— append-only log of every mention foundinstallation-prompt.md— install instructions for Claude AI and Zo Computerautomation-prompt.md— instructions for the hourly urgent-check + daily digest schedulestarter-prompts.md— example prompt for a first monitoring passassets/— pipeline diagram