Skip to content

docs(rig): restore the lost #368/#369 blocks, then move the verify queue out of tree - #421

Merged
mkovero merged 1 commit into
mainfrom
rig-queue-restore
Sep 1, 2026
Merged

mkovero merged 1 commit into
mainfrom
rig-queue-restore

Conversation

@mkovero

@mkovero mkovero commented Sep 1, 2026

Copy link
Copy Markdown
Owner

Two commits. The first repairs a loss; the second removes the mechanism that caused it.

1. Restore the blocks lost in a docs refresh

Rig-verification blocks written by the QA rounds on PRs #384 and #388, destroyed six days later. Both open QA findings currently have no recorded way to close them.

block added by lost in branch
#368 TAU_SNR_THRESHOLD_DB (17 lines) 8c36a060 dae1f47a issue-368
#369 xrun threshold (29 lines) d2cbf179 30f9c74a issue-369

Both chore commits are titled "refresh .agents, bin, .claude and rig from main" and both moved work/rig/rig-verify-queue.md to rig/ to match main's layout. The move was right; taking main's content along with the path was not. Git recorded each as a rename with edits (−35/+15 and −38/+15), which reads as tidy-up in the log and conflicts with nothing — so neither branch's queued measurement survived, and nothing flagged the loss.

Restored verbatim, placed together in Still to run ahead of the #243 AC7 bullet. Criteria and falsifying values unchanged. Each gains one line naming which branch to build: both describe fields main does not emit (tau_pre_impulse_snr_db, tau_reading{1,2}_xruns), so an operator on main would find nothing to measure and could read that as a pass.

2. Move the queue to $AC_HOME

The queue is per-rig coordination state, not per-branch state. In the repo, every branch got a copy and any branch could silently overwrite it. Moved to $AC_HOME/rig-verify-queue.md beside handoff/ and session/, which are out of tree for the same reason.

Content byte-identical to the deleted file (sha256 d8d651c9…af7a55), restored blocks included. Every in-tree reference repointed: TESTING.md, .agents/qa.md, .agents/rig.md ×3, CLAUDE.md, audit/rig-session-3/negative-lag-rule.md, work/rig/rig-verify-125-results.md, work/planning/usable-ir-path-issues.md ×2.

Two of those cited rig-verify-queue.md:339-366. Line numbers cannot survive a file with no revisions to diff, and they were already wrong — the block they point at was deleted and restored at a different offset. Both now name the block.

What this costs

Nothing versions $AC_HOME. The restore in commit 1 was only possible because the file was in git; after this, a deleted entry is gone with no git show to recover it. .agents/rig.md and CLAUDE.md both say so, and the role note now says to mark blocks executed rather than remove them. Ordering the commits this way keeps the loss and the restore in main's history permanently, after the file itself leaves the tree.

If that trade is worth tightening later, git init in $AC_HOME would restore recoverability without moving the file back.

Docs only. No code, no test, no wire format touched.

🤖 Generated with Claude Code

https://claude.ai/code/session_017tb982HsRpbLenwEq2cay1

…fresh

Both blocks were written by their PRs' QA rounds and destroyed six days
later by a chore commit, so two open QA findings currently have no
recorded way to close them.

- #368's `TAU_SNR_THRESHOLD_DB` block (17 lines) was added by 8c36a06
  and lost in dae1f47 on `issue-368`.
- #369's xrun-threshold block (29 lines) was added by d2cbf17 and lost
  in 30f9c74 on `issue-369`.

Both chore commits are titled "refresh .agents, bin, .claude and rig from
main" and both moved `work/rig/rig-verify-queue.md` to `rig/` to match
main's layout. The move itself was right. Taking main's *content* along
with the path was not: git recorded it as a rename with edits (-35/+15
and -38/+15), which reads as tidy-up in the log and conflicts with
nothing, so neither branch's own queued measurement survived and nothing
flagged the loss.

Restored verbatim from those two commits, placed together in "Still to
run" ahead of the #243 AC7 bullet, where the #369 block originally sat.
The criteria and their falsifying values are unchanged.

One line added to each that was not in the original: which branch to
build. Both blocks describe behaviour that exists only on their PR's
branch — `tau_pre_impulse_snr_db` and `tau_reading{1,2}_xruns` are fields
neither of which `main` emits — so a rig operator working from `main`
would find nothing to measure and could easily read that as a passing
run.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017tb982HsRpbLenwEq2cay1
@mkovero
mkovero merged commit 7511c39 into main Sep 1, 2026
@mkovero
mkovero deleted the rig-queue-restore branch September 1, 2026 11:27
@mkovero mkovero changed the title docs(rig): restore the #368 and #369 verification blocks lost in a refresh docs(rig): restore the lost #368/#369 blocks, then move the verify queue out of tree Sep 1, 2026
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