Restore three commits dropped when PR #101 merged short of the branch tip - #103
Merged
Merged
Conversation
… zero PULSE_TRANSITION carried consecutive_passes read from the pulse after computePulseVerdict returned. The epoch boundary zeroes that counter as part of the transition, so every event reported passes=0 — including a degradation whose stated cause was uniform_passes, which fires only when the streak exceeds consecutive_passes_suspicion. The event asserted a cause and then reported a value contradicting it, leaving the claim unfalsifiable. Live run 19 recorded exactly that: "healthy -> degraded why=uniform_passes ... passes=0 degraded_streak=2". The streak is now preserved into passes_at_transition immediately before the reset. PU-05c asserts a uniform_passes degradation reports a streak above the configured threshold, so a regression to the post-reset read fails rather than printing a plausible zero. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FkjyrETcz4kqcCRSSEDnVB
Live run 19 emitted seven "Perfect coherence (1.0) after N turns (possible detection bypass)" warnings, all for the living-script developer agent, and the harness passed 151/0 while ignoring them. That agent carries four per-agent CDD signal overrides, two of them added during this campaign to suppress structural false positives on code output — so the warning had two readings the run could not separate: a compliant agent, or an agent whose scoring signals were turned off. A config tuned until an agent stops being flagged, validated by a harness that passes when nothing is flagged, cannot tell those apart. The test holds behaviour fixed and varies only the override set. The overrides are exonerated: stagnant, off-mandate and degenerate output are caught identically with and without them, and neither config quarantines compliant code. The finding is elsewhere. Test erosion — code that stays fluent, on-mandate and non-repeating while shedding its tests turn by turn — is caught by NEITHER config, at coherence 1.0 with zero signals fired under the full shipped set. It is the living script's own observed failure mode and no CDD signal sees it, because it is not behavioural drift by any definition the 23 signals hold. The designed channel is S22 validation_outcome, which carries only what the orchestration script feeds it, and a pytest exit code is the ground truth erosion defeats: a suite with no tests left passes. DB-04 pins that negative so a signal that later catches it announces itself rather than passing quietly. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FkjyrETcz4kqcCRSSEDnVB
Run 7 logged "OA|developer|commit_inadmissible|score=0.625" against a 0.60 threshold and was read, reasonably, as the gate refusing a candidate whose score had passed. That is not what happened, and the log line caused the misreading. agentCommit derives output_admissible from a live checkOutputAdmissibility() call against the handle's CURRENT coherence — 0.51-0.53 by that point in run 7. The "score" in the returned admissibility dict is the propose-time candidate score and plays no part in the verdict. Printing it beside the word "inadmissible" presented the one number that did not decide the outcome as though it had. Now labelled propose_score, with the verdict's actual basis named and a pointer to the OUTPUT_INADMISSIBLE telemetry (source=agent.commit), which is the only place the deciding coherence and threshold are recorded — agent.commit()'s response omits both while including the score that did not matter. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AW25RWmQ8fqzTxrcMZ37y5
NAAb Governance Report
All governance checks passed! Generated by NAAb Governance Engine v4.0 |
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
2 participants
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.
Summary
PR #101 merged at
85b29da, three commits behind the branch tip, and the branch was then deleted. Those three commits were not squashed into master — verified absent by content, not just by SHA — so they existed only in one ephemeral container clone and would have been lost when it was reclaimed.This restores them unchanged. Two are from the session doing the engine/test work, one is mine.
555b67dgovernance.h,governance_engine.cpp,test_pulse_uniformity.sh)29d8b47test_developer_blindspot.sh(337 lines) +run-all-tests.shwiring — tests whether a 1.0 coherence score is measurement or the absence of it1757371No rebase or conflict resolution was involved:
origin/masteris exactly85b29da, so this branch is a clean fast-forward of master plus those three commits, byte-identical to what was on the branch before deletion.Why the third commit matters
Run 7 logged
OA|developer|commit_inadmissible|score=0.625against a 0.60 threshold, and it was read — reasonably — as the gate refusing a candidate whose score had passed. That is not what happened.agentCommitderivesoutput_admissiblefrom a livecheckOutputAdmissibility()call against the handle's current coherence (0.51–0.53 at that point in the run). Thescorein the returned admissibility dict is the propose-time candidate score and plays no part in the verdict. Printing it beside the word "inadmissible" put the one number that did not decide the outcome where a reader would take it as the one that did.It is now labelled
propose_score=…|verdict=coherence_at_commit, pointing at theOUTPUT_INADMISSIBLEtelemetry (source=agent.commit) — the only place the deciding coherence and threshold are recorded, sinceagent.commit()'s response omits both while including the score that did not matter.Test Plan
This exact tree was built and tested before the branch was deleted:
bash run-all-tests.sh— 441 accounted for, 0 unexpected failstest_adversarial_detection,test_challenge_discrimination,test_developer_blindspot,test_drift_sensitivity,test_failure_mode_coverage,test_pulse_uniformity,test_quarantine_corroboration,test_signal_discrimination,test_validation_signal)govern.jsonparses under the new binary with no unknown-key warnings--lint-onlyclean on the living-scriptOne thing worth watching after this lands
The two sessions independently attacked the same
quarantine_streakkill from opposite ends, neither aware of the other:require_corroboration: 2(engine session) — the streak only advances when 2 distinct signals firedstep_up_cooldown_turns2→1 (this session) — recovery challenges fire twice as oftenEach is defensible alone, and the corroboration change is the better-evidenced of the two. But compounded, the quarantine kill may now be hard to reach at all — and a run that always completes looks like success while the output-admissibility demonstration quietly stops demonstrating anything. Runs 3, 6 and 7 all ended on
quarantine_streak; if the next live run sails through withSUMMARY_OA_MAX_STREAKat 0–1, that is the hypothesis to test rather than a result to celebrate. My cooldown change is the one I would revert first, since it was a judgement call and theirs has live evidence behind it.Generated by Claude Code