[BUG] Title: Claude gave confident false verification results, wasting 2 hours of user time
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
Body:
User asked Claude to verify that documentation (SCHEMA.md, CLAUDE.md, SPEC.md) matched the actual database schema and code.
Claude delegated to a subagent which did a shallow check and returned "zero mismatches." Claude passed this along as verified without checking.
User caught this was wrong. Claude re-did the audit but got key facts wrong — claimed action_proposals columns (offer_multiplier, weight, dollar_value, missing_info) were "never queried" when they are actively written and read by multiple services and components.
Claude said "all done, all three docs agree" when they didn't.
User had to catch each mistake and push Claude to re-verify, three rounds total.
Each round, Claude presented results with high confidence ("zero mismatches", "no issues found", "all agree now") when the work had not been properly done.
Impact: ~2 hours of user time wasted. User lost trust in all output because each "verified" claim turned out to be wrong.
Root cause: Claude repeatedly marked work as complete without doing thorough verification. Delegated to subagent without validating results. Presented incorrect findings with false confidence multiple times.
What Should Happen?
When i tell caluse ti rad and matcgh docs, i doers so corectlty. I'm not wasdtig kore tiomes in this shiot. Fix the damn thuing
Irts wasted literal weeks of my time on ths projec alone becauen itn "skims" instead or reads. it "assumes" instead of reads. irt wqasted my fucking tiome insteakd of doi g work.
Error Messages/Logs
Steps to Reproduce
Fuck thios shiot.
Claude Model
None
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
23
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
_No response_
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗