UNI Universal Natural Intelligence

Wiki · Hierarchical Active Inference

Post-Closure Guard Report

Hierarchical Active Inference · hierarchical-aif/reports/POST-CLOSURE-GUARD-REPORT.md @ b909801f3db4 (hierarchical-aif/motor-stack) — opens the published snapshot 8b4b5935bcba

How to read this page

A Plain and a Clear version of this page have not been written yet. What follows is the document itself.

Precise — the source document

This is the document. Rendered from the repository at the commit above, with nothing rewritten for the web. A gate re-renders it on every deploy and fails the build if a single byte differs.

Date: 2026-07-22 · Gate: H-AIF-G9 · Context: all four corrected B4 cells landed; freeze before commit. This report records the state of every mechanical gate at the freeze point.

1. Repository state

check result
all new work under hierarchical-aif/ YES — git status --short shows nothing outside hierarchical-aif/ (94 entries, all under it)
frozen audits drift NONE — git diff --stat -- audits/ empty; full 250-file sha256 recheck IDENTICAL to frozen-evidence-baseline.sha256
result artifacts missing a sha256 sidecar NONE — all six primary results + two probe outputs + two probe JSONs now carry sidecars; sha256sum -c verifies OK on all six primary results
a template cited as a result NONE — the two templates live under reports/templates/ with TEMPLATE, NOT A RESULT banners; the only match for that banner outside templates/ is a prose mention in RUN-LANDING-MODE-STATUS.md
a canonical report path containing <<FILL>> / template text NONE — all four B4C##-CORRECTED-FULL-REPORT.md verified to be real reports, not templates

2. Guard results

gate command result
test suite python -m pytest hierarchical-aif/tests/motor_stack_aif -q 501 passed, 2 skipped, 1 xfailed
claim guard python …/claim_guard.py reports docs ledgers protocols scripts results 0 violations across 6 paths
numeric provenance guard python …/numeric_provenance_guard.py reports docs ledgers protocols 2046 in-scope decimals, 0 failures (2 ANCHORED_OK, 4 RECOMPUTED, rest MARKED/UNANCHORED)
frozen evidence 250-file sha256 recheck vs baseline IDENTICAL — NO DRIFT (receipt frozen-evidence-recheck-post-closure.sha256)

3. Result-hash sidecar verification

B4C01_CORRECTED_FULL_RESULT.json          : OK   564-… no, 8256cb12…
B4C02_CORRECTED_FULL_RESULT.json          : OK   0633988d…
B4C10_CORRECTED_FULL_RESULT.json          : OK   959a00e9…
B4C11_CORRECTED_FULL_RESULT.json          : OK   564a5b0f…
F_SIDE_MOTOR_STACK_SCORING_RESULT.json    : OK   b3b12720…
M4_M6_M7_PER_MOTOR_CONTRASTS_RESULT.json  : OK   751a59ef…

All six verified against their sidecars from the repo root (sha256sum -c returned OK).

4. The two mechanical guards, and what each cannot do

  • claim_guard clamps forbidden wording, use/mention aware. It cannot check whether evidence supports a claim, and it cannot check a number. Passing it is necessary, not sufficient.
  • numeric_provenance_guard (D11) checks declared numeric provenance: an anchored number must match the artifact field it names. It does not make numeric provenance decidable in general — UNANCHORED numbers are reported, not failed. The failing class is ANCHORED_MISMATCH.

Neither guard replaces the adversarial reader. Both are recorded with their scope so that a green guard is never mistaken for full coverage.

5. Verdict

All gates green. No frozen drift. No out-of-scope change. Every result hashed and verified. No template cited as a result. The working tree is in a durable, reviewable state and is ready for the principal's commit decision.

NEXT_ACT = write hierarchical-aif/reports/CORRECTED-CELL-RESULT-INDEX.md

sha256 8ac288dec0e1ee79 — of the original file, so what was ingested stays checkable.