Automation receipt smoke checks (2026-11-02)

[robgruen]

This is a dedicated, owner-authorized live smoke test for the completion receipts introduced in #506, following the merge of #507.

The check will request one /typechat draft, inspect the fixed completion receipt, and rerun only the receipt controller to verify that it does not duplicate comments or invoke another model run.

This is not a feature proposal, work-acceptance request, coding-run authorization, or merge decision. The coding pilot remains disabled. No model draft will be published or accepted. Maintainers will distinguish actual live results from offline tests and any checks still pending.

[robgruen]

/typechat draft

[github-actions[bot]]

TypeChat: Draft prepared; owner or maintainer input is still needed before work acceptance.

Owner request | Run and preview

Operational receipt only. This comment does not publish draft text, accept work, authorize implementation, or approve merging.

[robgruen]

Owner-authorized receipt verification results

The approved live checks passed on revision a36fe49cc3ab74b46dd752e075c918b52e801ff8.

Check	Observed result	Evidence
Issue draft receipt	Eyes acknowledgment, then exactly one canonical completion receipt; no draft text published	Issue receipt, source run
Discussion draft receipt	Eyes acknowledgment, then exactly one canonical completion receipt	Discussion receipt, source run
Invalid-command failure	Rejected during preparation; generation, detection, sealing, and public-reply jobs skipped; one fixed error receipt posted	Error receipt, expected-failure run

All three receipt-controller retries succeeded without another comment or mutation:

• Issue controller, attempt 2
• Discussion controller, attempt 2
• Error controller, attempt 2

Receipt IDs, text hashes, update timestamps, and comment counts stayed unchanged across those retries. New-intent, comment-write, and result-upload steps were skipped. Every source run remained at attempt 1; only the controllers were rerun.

Both valid drafts retained their exact frozen-context hashes after their operational receipts and retries. The later invalid command intentionally added new Issue evidence; it was not treated as an approval or a reason to reuse an old preview. The existing #504 acceptance record remained unchanged.

Native usage summaries reported 61.35286 AI credits total: Issue drafting 33.70223 plus detection 6.24782; Discussion drafting 12.89179 plus detection 8.51102. This was within the authorized maximum of 900. Retries and the invalid-command check executed no inference.

Limits: these live checks covered draft outcomes reporting that input is still needed, successful no-op controller retries, and invalid syntax. Cancellation, uncertain-network-write recovery, acceptance-ready receipt wording, and direct-reply reuse were not exercised live here. Offline tests are not being presented as live evidence for those paths.

No draft was published or accepted, no issue was closed, and no coding run or coding-budget activation occurred. The implementation activation flag remains unset at repository and inherited organization scope. A first coding pilot still requires a separate owner decision and genuinely accepted, supported work without a competing contribution.

Comment-count assertions above were completed before this maintainer-facing results comment.