docs: report the B01 pilot and what its checks caught - #424
Merged
Merged
Conversation
B01 ran on 2026-09-20; the fix merged as c4030a4 (#421, closing #412). All six preregistered acceptance checks are satisfied. The report states plainly what this was: a host agent executing the contract's checks, NOT a CodeCartographer engine running a change. E02 storage and E05/E06 evidence collection do not exist, so no record was minted, no digest bound, no acceptance classified. It measures whether the checks are usable and load-bearing. The bug also had a public diagnosis in its issue, which makes it a known-answer protocol test rather than evidence of unaided diagnosis; the rubric requires reporting that exposure. What the checks caught that the implementer did not: - A6's mandated independent review found the fix incomplete. Git has four environment-reachable config sources; the issue named one gap and the fix closed it, leaving GIT_CONFIG_PARAMETERS live. The original failure reproduced byte-for-byte on a branch with green CI whose author believed it done. - A3's fresh-child-process requirement is why the tests could not be faked by an earlier test's environment cleanup. - A1's before-the-fix ordering produced a negative control that later proved the target test was never vacuous. - A6 also caught a test that resisted its own fix: it asserted the helper sets exactly three variables, so covering a fourth source broke it. Two claims made during implementation were wrong and are corrected in the report rather than left standing: the guard's old position was a latent hazard, not a realized defect (the reviewer ran that file under injection: 10/10 passed), and git reads an empty GIT_CONFIG_PARAMETERS as zero entries rather than erroring. The report also names a gap the pilot exposed, filed as #422: the checks ask whether the reported defect is fixed and proven, never what else of the same class exists. Three review rounds hit that pattern. #423 records the GIT_TEMPLATE_DIR follow-up left out of scope.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Records the first real change run through the engineering contract's preregistered acceptance checks. Bug fixed and merged as
c4030a4(#421, closing #412); all six checks satisfied.What the report says, and refuses to say
It states up front that this was a host agent executing the contract's checks, not a CodeCartographer engine running a change — E02 storage and E05/E06 evidence collection do not exist, so no record was minted, no digest bound, no acceptance classified. It measures whether the checks are usable and load-bearing, nothing more. It also reports that #412 carried a public diagnosis and suggested fix, making this a known-answer protocol test rather than evidence of unaided diagnosis, as the rubric requires.
The finding worth keeping
A6's mandated independent review found a blocking defect in a change with green CI whose author believed it complete. Git reads configuration from four environment-reachable sources; the issue named one gap, the fix closed it, and
GIT_CONFIG_PARAMETERSstayed live. The original #412 failure reproduced byte-for-byte through it.Three other checks earned their place: A3's fresh-child-process requirement made the tests unfakeable by environment cleanup; A1's before-the-fix ordering produced a negative control that later proved the target test was never vacuous; and A6 additionally caught a test whose shape penalized widening the isolation it existed to protect.
Corrections to its own claims
Two statements made during implementation were wrong and are corrected rather than left standing — the guard's old position was a latent hazard rather than a realized defect (10/10 passed under injection), and git reads an empty
GIT_CONFIG_PARAMETERSas zero entries rather than erroring. A contract that produces confident-sounding unverified claims is worse than one producing none.Gap filed, not buried
#422 — the checks never ask what else is of the same class. Three review rounds hit that pattern: E03 round 1 fixed
repositoryand leftexecutablecoerced; B01 fixedGIT_CONFIG_COUNTand leftGIT_CONFIG_PARAMETERS.#423 —
GIT_TEMPLATE_DIRfollow-up, deliberately left out of #421's scope.Docs only.