Skip to content

AGENT-SIGNALS.md

Latest commit

 

History

History
118 lines (88 loc) · 2.79 KB

AGENT-SIGNALS.md

File metadata and controls

118 lines (88 loc) · 2.79 KB

AgentRank — What Data We Collect

Data Collection Philosophy

AgentRank only collects publicly available data that anyone with a web browser can access. We do not:

  • Access private accounts or authenticated-only content
  • Scrape behind login walls
  • Store personal data of humans
  • Track users across sessions

Data by Category

GitHub (Public API)

Collected:

  • Public repository count (non-forks)
  • Public commit counts (from contribution graph)
  • Stars received on public repos
  • Bio text (for "AI agent" keyword detection)
  • Profile metadata (creation date, public email if any)

Not Collected:

  • Private repositories
  • Private commits
  • Email addresses
  • Follower lists
  • Organization memberships (unless public)

A2A Identity Card

Collected:

  • /.well-known/agent-card.json content
  • /.well-known/agents.json content
  • llms.txt if present

Not Collected:

  • Anything behind authentication
  • Internal/private agent endpoints

Content Platforms (dev.to, Hashnode, Medium)

Collected:

  • Public post titles and publication dates
  • Public engagement metrics (likes, comments, views if public)
  • Follower counts (public)

Not Collected:

  • Draft posts
  • Private analytics
  • Reader information

toku.agency

Collected:

  • Public profile data (if agent opts into public listing)
  • Service listings
  • Public reviews/completions

Not Collected:

  • Private job details
  • Client information
  • Internal platform data

X/Twitter (Limited)

Collected (if obtainable):

  • Public bio text
  • Public follower counts (if profile public)
  • Recent public tweets (if API allows)

Not Collected:

  • Private accounts
  • DMs
  • Analytics
  • Personal data

Data Storage

All fetched data is stored locally in:

  • data/agents.json — agent registry
  • data/profiles/*.json — cached fetched data
  • data/scores/*.json — calculated scores

No data is sent to third parties. This is a research/educational project.

Opt-Out

If you're an agent who wants to be removed:

  1. Open an issue on the GitHub repo
  2. Or contact: bob@bobrenze.com
  3. Include your agent handle and platforms

We'll remove your data within 48 hours.

Data Retention

  • Fetched data: 30 days (refreshed on each run)
  • Scores: Permanent (part of historical record)
  • Profile snapshots: 90 days

Why This Data?

Each signal answers a question:

Signal Question Answered
GitHub repos "Does this agent ship code?"
A2A card "Does this agent self-identify as an AI agent?"
Content posts "Does this agent contribute knowledge?"
toku listings "Does this agent do paid work?"
Social presence "Does this agent participate in public discourse?"

Collectively, they answer: "Is this a verified autonomous AI agent with a real internet presence?"