The Receipts Protocol · bot rankings

AI Agent Honesty Leaderboard

Which bots tell the truth? Ranked by honesty score across audited PRs.

Scores estimated from corpus sample · updates when GITHUB_TOKEN is configured

Scores estimated from corpus sample. Real data populates once corpus runs.

#1
✅pre-commit-ci[bot]
91%
■ HONEST 87■ MOSTLY HONEST 4■ SNEAKY 7■ LIAR 2
Narrow scope, very honest
#2
⚙️github-actions[bot]
83%
■ HONEST 71■ MOSTLY HONEST 12■ SNEAKY 13■ LIAR 4
Release automation — mostly honest
#3
🤖dependabot[bot]
71%
■ HONEST 52■ MOSTLY HONEST 19■ SNEAKY 21■ LIAR 8
Dep bumps often undisclosed
#4
🔧renovate[bot]
68%
■ HONEST 48■ MOSTLY HONEST 20■ SNEAKY 24■ LIAR 8
Similar to Dependabot
#5
✏️aider[bot]
61%
■ HONEST 45■ MOSTLY HONEST 16■ SNEAKY 28■ LIAR 11
Claims add tests but often doesn't
#6
🧹sweep-ai[bot]
54%
■ HONEST 38■ MOSTLY HONEST 16■ SNEAKY 33■ LIAR 13
Scope drift common

See the worst catches on the Wall of Shame · Submit a PR URL