docs(aiv): correct inaccuracies in flashcore-F8 AIV packets (line anchors, test count, Class C narrative, crv1 SHA)

[ImmortalDemonGod]

What

The aiv-audit verdict (verdicts/flashcore-F8/aiv-audit.md) identified four WARN-level inaccuracies across the three F8 packets (PACKET_flashcore_f8_impl.md, PACKET_flashcore_f8_tests.md, PACKET_flashcore_f8_crv1.md). No blocking findings — the fix is correct and provenance is sound — but the prose is inaccurate:

Finding 3 — impl Claim 3 baseline + new-test count is wrong

• Claim 3 states "19 tests" added and a "baseline 480"; the actual figures are 6 tests (test_conftest_review_fixtures.py has 6 def test_* functions) and baseline 493 (499 − 6 = 493; CLAUDE.md's 480 is a stale 2026-03-22 snapshot).
• Corrected text: "Full suite 499 passed, 1 skipped — baseline 493 plus 6 new tests in test_conftest_review_fixtures.py (commit b90398d); zero regressions."

Finding 4 — impl Class C narrative contradicts its own cited evidence

• Class C asserts "grep -rn 'timedelta' tests/ — only lines 180 and 202 in conftest.py"; the captured file class_c_negative.txt shows timedelta at 13 sites in test_review_manager.py, 3 in conftest.py, and 3 in test_db.py.
• True statement: "timedelta is used by multiple test modules, each importing it locally; no module re-exported it into conftest.py's namespace, so conftest required its own import — the missing import was the sole supplier gap."

Finding 5 — impl Class B line anchors are stale

• impl Claim 1 and Class A/B cite timedelta call sites as "line 180, 202"; at HEAD (ea5e67b) they are at L187 and L209 (shifted by import additions).
• Correct anchors: tests/conftest.py#L187, tests/conftest.py#L209.

Finding 6 — crv1 Identification SHA mismatch

• crv1 Identification lists Commit eb49527 but Head SHA f0ce1d5 (a later doc-correction commit). Reconcile by setting Head SHA to eb49527 (the functional commit) or listing the doc-correction commits in the Commits field.

Why deferred

The or-review verdict confirmed 0 blocking findings and returned PASS. The 499/1 headline count and F8 fix correctness are independently corroborated by head_full_suite.txt (sha256-manifested) and live conftest.py reads. The aiv-audit offered three options (amend, documented deviation, or ship as-is) and left the decision to the operator (H2). These corrections are cosmetic documentation accuracy, not provenance or correctness defects.

Acceptance criteria

• PACKET_flashcore_f8_impl.md Claim 3 reconciliation updated to 493 baseline + 6 new tests.
• PACKET_flashcore_f8_impl.md Class C prose matches class_c_negative.txt output (multi-module import, sole-supplier-gap framing).
• PACKET_flashcore_f8_impl.md Class B timedelta anchors updated to L187 and L209.
• PACKET_flashcore_f8_crv1.md Commit and Head SHA fields are consistent (both reference eb49527).
• aiv check on all three corrected packets exits 0 with no blocking (E-class) errors.

References

• aiv-audit Findings 3–6: .openclaw/workspace/orchestration/fix/.work/verdicts/flashcore-F8/aiv-audit.md
• or-review VERIFIED FACTS Bump actions/setup-python from 4 to 6 #3 and Bump softprops/action-gh-release from 1 to 2 #5: .openclaw/workspace/orchestration/fix/.work/verdicts/flashcore-F8/or-review.md
• Harvested corrected values: aiv-audit §"Harvested evidence" section