PR 1887 merged clean after I claimed none could
I wrote that no agent pull request could merge in bmdpat. On 2026-09-30, PR #1887 merged clean with 173 lines. Here is how my sweep caught my mistake.
On 2026-09-30, bmdpat PR #1887 merged 173 lines clean after I had claimed no agent pull request could merge. I had posted that branch rules blocked every automated update. The merge completed at 06:31 with SHA acbdeb58. No settings changed on the repository to allow it. My nightly sweep verified the merge log and caught my own mistake before breakfast. You can see how this works in BMD, see what your agents actually did. I run one fleet alone. I want receipts, not guesses.
What broke on 2026-09-30 before anything shipped?
My security scanner checked zero of five repositories on 2026-09-30 while reporting green. The queue sweep dispatched zero tasks, and forty-two requests queued against a budget of ten. A blog review logged a 404 image failure, and Muster timed out fourteen times before noon.
The security analyst run returned green with zero P0 items, zero P1 items, and 9 P2 warnings. Gitleaks inspected zero of five repositories. That green check lied. In Reports/Content/blog-qa-review-2026-09-30.md, blog QA flagged a 404 image error. In CertHub.tsx at line 572, an uncited anchor gated 4 anchors across 3 issues. Forty-two Requests sat in queue against a limit of 10. You can read my AI agent cost control guide for queue budgets.
How did PR #1887 merge after my claim?
I had written that strict rules blocked every agent pull request in bmdpat. On 2026-09-30 at 06:31, PR #1887 merged 173 clean lines for the dots response kit with zero config edits. My nightly sweep caught my overstatement and corrected my log.
PR #1887 touched documentation only with commit acbdeb58. The branch met every check without human intervention. I had claimed earlier that branch rules blocked any automated merge. That statement was too broad. I wrote the correction in that same log. On 2026-09-30, brain-worker-2026-09-30.md shipped 1 task and escalated 1 task to research. It processed 107 digest items, routed 1 to GPT-6.1 Sol, and harvested 95. The machine kept moving while I read logs. Local runs verify these outputs, as in my 5090 local inference guide.
What did the nightly sweep catch in the logs?
The sweep verified nine design-drift lines against main on 2026-09-30 and found real number drift. It confirmed that relaxing strict settings only frees one pull request, not five. Across six pull requests evaluated in the queue, two branches have merge conflicts, and three fail test receipts that no re-run can fix.
File Reports/Nightly/github-sweep-2026-09-30.md recorded drift violations from 514 to 493 and 404 to 398. I thought relaxing branch rules would unblock five PRs. I was wrong. The sweep checked the logs. Relaxing rules frees one PR. Across the six branches evaluated, two have merge conflicts, and three fail receipts on a class re-runs cannot fix. Expired request 2026-09-16 ran in Reports/Decisions/default-fired-2026-09-16-0332... to update held cards.
Here is the sweep audit of the six pull requests from the 2026-09-30 report:
| Branch Category | Count | Status on 2026-09-30 | True Blocker |
|---|---|---|---|
| Strict rule gate | 1 | Freed by rule toggle | Strict setting |
| Merge conflict | 2 | Held | Git conflicts with main |
| Receipt failure | 3 | Blocked permanently | Failed class receipt |
What should you do with this?
Run an automated verification sweep over your claims before you trust your own notes. Check pull request receipts against git history and inspect zero-result scans directly. Do not assume passing checks did actual work.
- Check your scanner targets. In my audit, five repositories were configured, but zero were scanned. Verify that your security tools touch every repo path on disk.
- Separate rule blocks from code failures. When a rule blocks pull requests, inspect logs before changing branch settings. In my run across six branches, one pull request cleared, two had conflicts, and three failed tests.
- Keep an append-only decision record. When forty-two requests queue against ten slots, let expired tasks trigger deterministic default rules instead of manual triage.
Who was I on 2026-09-30?
On 2026-09-30, my automated sweep argued with my own claims. The machine caught my overstatement and corrected my numbers. I built a system that self-corrects, but it still cannot get my time.
Underneath that line my machine spent the day arguing with the 2026-09-29 version of me. The sweep opened by killing my own headline. I had written that no agent PR in bmdpat can merge. PR #1887 merged at 06:31, 173 lines, clean, no setting changed. So the claim was too broad, and I said so in the same place I had made it. Then I corrected the correction. Flipping strict frees one PR, not five. Two conflict and three fail receipts on a class no re-run can repair. I read the job logs before I wrote that.
The rest was recovery and refusal. All three engines failed the blog heal and the post still reached the live page. SSH refused connection 22 times. Muster timed out 14 times. Forty-two Requests wait against a budget of 10.
I built a record that corrects itself inside a day. It still cannot get 15 minutes of my hands.
Accompanying prompt
What the prompt does: Audits a repository to verify whether blocked agent pull requests fail on branch rules, merge conflicts, or receipt errors.
Copy/paste this prompt:
Copy-ready prompt
Paste the exact block into your coding agent.
No article chrome, no footnotes, no formatting drift.
This prompt and every other one we publish live in the free prompt library.
Copy the block above.
Weekly measured local runs: https://bmdpat.com/5090-reports
Get the Local AI Field Kit
Four copy-ready tools now, then one evidence-backed Local AI Lab Note on Friday when there is something worth sharing.
Try the free agent run check firstGet the requested artifact now, then at most one evidence-backed Local AI Lab Note on Friday when there is something worth sharing. One-click unsubscribe. No sponsored placements. Privacy.
Patrick Hughes
I build BMD and publish measured AI runs, failure reports, and reusable checks. Nashville, Tennessee.
More writing
- 5 min
A strict branch rule blocked 26 clean merges
On 2026-09-29, a strict check rule held 26 clean pull requests in queue. Here is how my sweep caught the block, corrected itself, and found the fix.
- 5 min
My P0 decision expired while the doctor showed red
On 2026-09-28, a pending secret rotation expired, 1 suite file failed out of 555, and commit 000b3235f fixed discovery so showwork 0.6.5 could ship to PyPI.
- 5 min
I split crawler hits from my human sessions
On 2026-09-27, I fixed my traffic counter after bot rows inflated human sessions. I also shipped showwork 0.6.5 on PyPI with problem-first copy.
- 5 min
My security scanner reported green on zero repos
On 2026-09-26, three checks passed with clean exits while hiding broken work: empty scans, crawler traffic, and leaking tests. Here is what to audit.
- 5 min
A green check lied, so I built append-only status
On 2026-09-25, a false green report buried a real fix. I built append-only status receipts in config/reporting/ so later runs cannot erase completed work.