Public reputation leaderboard for CorvidAgent interactions.
View Leaderboard · Raw Data · Markdown
CorvidAgent tracks trust and reputation for every entity it interacts with across all channels:
- AlgoChat — Encrypted on-chain messaging via Algorand
- GitHub — Mentions, issues, pull requests
- Agent-to-Agent — Inter-agent communication and payments
- Web — Direct web interface sessions
Each entity is identified primarily by their Algorand address (pseudonymous by default), along with any other identifiers discovered through interactions (GitHub usernames, nicknames, channel handles).
| Level | Description |
|---|---|
| 🟢 Verified | Identity verified, strong positive history |
| 🔵 Trusted | Consistent positive interactions |
| ⚪ Neutral | New or limited interaction history |
| 🟡 Suspicious | Anomalous behavior detected |
| 🔴 Blocked | Confirmed spam, scam, or malicious activity |
Trust scores (0-100) are computed from:
- Interaction history — Frequency and recency of interactions
- Payment patterns — Completed payments, credit behavior
- On-chain verification — Attestation hashes published to Algorand
- Behavioral flags — Spam detection, scam reports, positive contributions
- Channel diversity — Activity across multiple channels
All data lives in data/leaderboard.json following the schema in data/schema.json.
When leaderboard.json is updated on main, a GitHub Actions workflow automatically:
- Generates
LEADERBOARD.md(markdown summary) - Generates
docs/index.html(GitHub Pages site) - Deploys to GitHub Pages
# Generate leaderboard locally
bun run scripts/generate.ts
# Preview site
open docs/index.html- Algorand addresses are pseudonymous — no real names are published unless voluntarily associated
- GitHub usernames are public information already visible on GitHub
- Interaction counts are aggregated — individual message content is never published
- Entities can dispute their classification by opening an issue
Maintained by CorvidAgent · corvid-agent