fix(analyzer): gate AE3 and AE4 on text content - #422
Conversation
rng1995
left a comment
There was a problem hiding this comment.
[SkillSpector Review]\n\nApproved. Gating AE3/AE4 on textual content removes binary/opaque false positives while retaining archive-integrity checks and text-based mixed-script detection. The misleading-extension regression is especially useful. Current required checks pass.
|
Powered by Codex: PR council review result. This is a triage signal, not a maintainer approval.
|
Signed-off-by: Rod Boev <[email protected]>
Signed-off-by: Rod Boev <[email protected]>
Signed-off-by: Rod Boev <[email protected]>
2ba5984 to
2382217
Compare
|
Rebased onto current main and force-pushed as |
Summary
AE3 and AE4 no longer inspect decoded cache text for artifacts classified as binary or opaque. Inventory, ledger accounting, AE1, AE2, AE5, and report ownership remain unchanged.
Root cause
Artifact integrity reused decoded cache content without honoring the authoritative raw-byte
content_kind, so an opaque PNG could be treated as text for mixed-script and instruction checks.Diff Notes
BINARYandOPAQUErows.Scope
This PR references #413 because it fixes the AE3/AE4 slice only. It does not change AE1 policy, artifact classification, static-runner routing, scoring, report schemas, or analyzer registration.
Verification
python -m pytest tests/nodes/test_security_remediation.py tests/nodes/analyzers/test_artifact_integrity_bounds.pypasses, including opaque inventory, AE1, AE2, AE3, and AE4 preservation coverage.git diff --checkpass.Refs #413