Observed arrival · 2026-11-17
donediff, a receipt-checker for coding agents
A CLI audits Claude Code session transcripts by comparing an agent’s claims with recorded commands, results, and edits.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Claude Code users auditing agent-run sessions
- Worth noticing
- The sample flags `|| true`, a test reduced from two assertions to one, and a commit made with `--no-verify`.
Field notes
A session is scored by comparing extracted agent statements with transcript evidence and the ordering of commands and edits: a passing check only verifies a later claim if no code changed afterward. The tool also flags tactics such as masked command failures, reduced test assertions, and bypassed Git hooks. Local Git checks can confirm a named file or commit, but the page says donediff does not verify that the resulting code actually does what the agent described.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue