Skip to content

docs(adr): record the measured effect of the handoff change in ADR 0010 - #70

Merged
chrispezza merged 1 commit into
mainfrom
docs/handoff-trial-result
Oct 12, 2026
Merged

chrispezza merged 1 commit into
mainfrom
docs/handoff-trial-result

Conversation

@chrispezza

Copy link
Copy Markdown
Collaborator

Summary

Appends the result of the handoff trial to ADR 0010, replacing its earlier statement that nothing had measured the change. Docs only.

  • Old handoff: the agent built the unspecified is_favorite field in 6 of 6 runs and left the spec untouched.
  • New handoff: it withheld the field, proposed the spec change and asked in 6 of 6.
  • No halt on in-spec work in either arm, including four control runs.
  • The failure the issue predicted (a halt) did not occur. The real one was code and spec diverging.

Limits are in the amendment: one fixture, one model, one request, six runs per cell, no follow-up turn.

Test plan

  • python checks/run_checks.py --ci --require-validator — 0 findings, index regenerated
  • CI green

🤖 Generated with Claude Code

Sixteen isolated headless runs on the tea-tracker package. With the old
handoff the agent built an unspecified field in 6 of 6 runs and left the spec
untouched; with the new section it withheld the field, proposed the spec
change and asked in 6 of 6. No halts on in-spec work in either arm.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@chrispezza
chrispezza merged commit 6c463dc into main Oct 12, 2026
2 checks passed
@chrispezza
chrispezza deleted the docs/handoff-trial-result branch October 12, 2026 19:15
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant