session 2026-08-26: PENDING-163/164/165 + five amendments; prior-art search built

Session record: the record did not contain the decision.

Filed: PENDING-163 (+3 amendments), PENDING-164 (jurist-drafted, placed verbatim,
+ executor addendum and build record), PENDING-165 (+1 amendment). Rulings
REVIEWED-130 and REVIEWED-131 placed by the steward.

Built: preserve-transcripts.py (PENDING-147 (i), 43 transcripts read-back proved),
two pre-commit fixes, the hook-directory allowlist in governance-drift-check.py,
and prior-art.py + the prior_art MCP tool (PENDING-164 (c)+(d)).

The finding: a steward decision that rewrote seventeen commits of history
(chamber-library 0677e8a, retiring LFS) is absent from the entire authorization
record, so both AI parties independently recommended adopting the mechanism it
retired. governance_search over 313 items returns one hit; grep over the raw files
returns zero before today.

Six self-referential instrument failures and three false zeroes, one of which was
caught before it was believed — and the only reason was that the jurist
pre-specified what the instrument had to return.

MEMORY.md rotated (prior Active Session demoted verbatim, checked for carried
claims first). 6 KG lines incl. one genuine transfer: the 2026-08-25 lesson about
declaring a limit rather than manufacturing a column fired unprompted on a
different instrument in a different item. One skill-harvest proposal, whose subject
is that today's own build ignored the routing table measurement it had open.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NvZAKSf9aqratbqHbU9LK5

claude/memory/knowledge-graph.jsonl

@@ -731,3 +731,9 @@
 {"subject": "declaring a limit instead of manufacturing the column", "predicate": "prevention", "object": "STOPPED STEP 1 FROM BECOMING STEP 2 IN DISGUISE. PENDING-151's pre-registration asked for terms, entities AND propositions. Propositions cannot be extracted mechanically \u2014 it requires reading for claims \u2014 and the only reader available at step 1 was the party barred from judging. Building that column with a model would have been the forbidden judgement wearing the permitted step's clothes. Declaring it absent kept the routing intact, and the jurist then judged propositions itself from the raw pairs.", "valid_from": "2026-08-25", "valid_to": null, "confidence": 0.95, "source_file": "session-2026-08-25-tarbuckle-wired-and-the-contact-gap.md", "extracted_at": "2026-08-25"}
 {"subject": "the steward's mandated commensurability check", "predicate": "prevention", "object": "FOUND A COLLISION IN THE DIRECTION NOBODY WAS LOOKING. The check was ordered for any periodicity tier 3 introduced; a seam is aperiodic, so the answer looked trivially empty. Running it anyway surfaced that the mumble's clock file PERSISTS ACROSS SESSIONS, so any gap over the interval left a tick already due at the moment of waking \u2014 the fool would have spoken twice into the same seam. An instrument aimed at one thing caught another because it was actually run.", "valid_from": "2026-08-25", "valid_to": null, "confidence": 0.95, "source_file": "session-2026-08-25-tarbuckle-wired-and-the-contact-gap.md", "extracted_at": "2026-08-25"}
 {"subject": "chamber-v1 archive", "predicate": "has-property", "object": "9 complete formation pairs, 6 sessions, 3 protocol axes, 19479 words raw+submitted, 33 real content files. '55 files' is the AppleDouble-inclusive count and is unstable across days. Claude arm longer in 9 of 9 pairs, 1.41x-3.40x \u2014 formation and length are perfectly confounded, so the surplus half of PENDING-151's question is unanswerable from this corpus.", "valid_from": "2026-08-25", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-25-tarbuckle-wired-and-the-contact-gap.md", "extracted_at": "2026-08-25"}
+{"subject": "claude-code", "predicate": "drift-pattern", "object": "A POLICY QUESTION MANUFACTURED BY MY OWN ERROR. PENDING-163 went to the steward as a decision between two options because the item claimed option (ii) 'widens what may be committed everywhere'. It does not: git cat-file -s reads the STAGED blob, so an LFS-tracked file measures 133 bytes and a plain one still measures 17MB and is still refused. The claim was false by a category, it made one option look safer and the other costlier than either was, and the steward's time was spent on a fork that did not exist. Not a judgement call under uncertainty — a confident wrong claim that generated a decision.", "valid_from": "2026-08-27", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-26-the-record-did-not-contain-the-decision.md", "extracted_at": "2026-08-27"}
+{"subject": "claude-code", "predicate": "drift-pattern", "object": "DID NOT CHECK PRIOR ART, AND WAS THE ONLY PARTY WHO COULD. Both AI parties recommended adopting Git LFS twelve weeks after the steward had tried it, documented why it failed, and rewritten seventeen commits to undo it. The jurist COULD NOT have found it — repo_activity caps at 100 commits, five weeks short of 0677e8a — so 'search prior art before proposing' is not a rule it can follow. The executor has a filesystem and did not look until the string 'pre-lfs-export' appeared in an unrelated directory listing. The asymmetry is structural; the failure to use it was not.", "valid_from": "2026-08-27", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-26-the-record-did-not-contain-the-decision.md", "extracted_at": "2026-08-27"}
+{"subject": "claude-code", "predicate": "drift-pattern", "object": "THE CHECK AGAINST BLIND CHECKS WAS BLIND TO ITSELF. Five new controls for the hook-allowlist registered AFTER failed_controls is computed (line 846 vs 702): all ran, none counted, the tally still read 48/48, and a failure among them would have printed NOTHING. Built in the same hour as, and as the fix for, the class 'a check that passes because it could not run' — immediately after the jurist had caught the previous blindness in the same item. Caught only by comparing the printed tally against the number of controls I knew I had added.", "valid_from": "2026-08-27", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-26-the-record-did-not-contain-the-decision.md", "extracted_at": "2026-08-27"}
+{"subject": "claude-code", "predicate": "drift-pattern", "object": "THREE FALSE ZEROES IN ONE DAY, EACH FROM A DIFFERENT MECHANISM. (1) A zsh loop over a newline-joined string iterated once and printed 'NONE' having measured nothing. (2) sh() discarded stdout on non-zero exit, and find over $HOME exits 1 from unreadable Library dirs WHILE printing all 37 repos, so every prior-art search returned zero. (3) The uncounted controls above. A clean zero is the most dangerous output an instrument produces, because it is indistinguishable from a real negative and carries the same confidence.", "valid_from": "2026-08-27", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-26-the-record-did-not-contain-the-decision.md", "extracted_at": "2026-08-27"}
+{"subject": "the jurist pre-specifying what an instrument MUST return", "predicate": "prevention", "object": "STOPPED A FALSE ZERO FROM BEING BELIEVED — the only one of three that day caught before it was acted on. The jurist specified, before prior-art.py was written, that its positive control must return 0677e8a and 95760ff or it measured nothing. Its first run returned zero for every term (find exits 1 on unreadable dirs; stdout was being discarded), and the control failed loudly instead of the tool reporting 'no prior art' forever. The other two false zeroes that day were caught by luck — an echoed cd error, and a tally that looked one short. The difference between them is not care; it is that someone external named the expected output in advance.", "valid_from": "2026-08-27", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-26-the-record-did-not-contain-the-decision.md", "extracted_at": "2026-08-27"}
+{"subject": "declaring a limit instead of manufacturing the column (banked 2026-08-25, PENDING-151 step 1)", "predicate": "prevention", "object": "STOPPED A DIFFERENT INSTRUMENT FROM FAKING ITS SECOND HALF, TWELVE DAYS LATER AND IN ANOTHER ITEM. PENDING-164's census has two halves: grep adoption/retirement verbs (mechanical), then check each hit against the register (requires reading each commit to identify WHICH mechanism it decided about — interpretation). The verb grep returned 661, narrowed to 270. The lesson from PENDING-151 — where propositions could not be extracted mechanically and the column was declared absent rather than built with a model — fired here unprompted: the census reports 270 UNCLASSIFIED candidates and states that the backlog is not censused. This is transfer, not repetition: different item, different instrument, same shape.", "valid_from": "2026-08-27", "valid_to": null, "confidence": 0.95, "source_file": "session-2026-08-26-the-record-did-not-contain-the-decision.md", "extracted_at": "2026-08-27"}