Skip to content

Explain disclosure holds and review the October 8 candidates - #13

Open
JunjoSick wants to merge 1 commit into
mainfrom
fix/disclosure-review-diagnostics
Open

JunjoSick wants to merge 1 commit into
mainfrom
fix/disclosure-review-diagnostics

Conversation

@JunjoSick

@JunjoSick JunjoSick commented Oct 9, 2026 •

Copy link
Copy Markdown
Owner

The October 8 daily run used ambiguous-disclosure for every recent non-blocked article, including articles with no advertising disclosure. That made its 77-row queue look more suspicious than the underlying evidence.

This PR adds precise review diagnostics and documents the completed 77-path investigation:

  • GDC: 5 publisher-disclosed advertisements, recommended for a separate reviewed exact-path addition.
  • QuiAntella: 27 ordinary editorial articles and 2 unresolved retail-opening features; no confirmed ads.
  • Colli: 42 ordinary editorial articles and 1 unavailable article; no confirmed ads.

All 77 paths are absent from the current registries. Current RSS/API fragments were recovered for 76 candidates. Old and new classifiers returned identical decisions for all 76. GDC has three overlapping causes: styled PRE labels on all five, linked attribution footers on all five, and gallery STYLE elements on three.

Changes:

  • Report missing supported disclosure, active content, unsupported label context, and unsafe following content separately.
  • Forward diagnostics through the existing worker and discovery report.
  • Add invented regression cases, including the layered GDC shape, hidden/quoted evidence, 599/600-character boundary, later valid labels, and byte-stable publication outputs for held candidates.
  • Include primary URLs, per-candidate assessments, source fingerprints, limitations, and narrowly scoped follow-up recommendations in the review report.

The accepted disclosure predicate, vocabulary, active-content veto, resource limits, timestamps, conflicts, and publisher gates are preserved. No baseline/state/registry/filter changes; no automatic publication of the recommendations. The 69 editorial assessments are evidence-based descriptions, not permanent keep overrides or proof that undisclosed payment never occurred.

Validation:

  • Focused article-card tests: 30/30 passed.
  • Current-body comparison: 76/76 unchanged decisions.
  • npm run cards:check, npm run check, bundle history verification and whitespace checks passed.
  • Full local suite: 125 passed; the required browser test could not run because Chromium is unavailable locally. The attempted Playwright download was invalid.
  • GitHub CI passed, including the mandatory Chromium/browser lifecycle test and generated-output/history checks: https://github.com/JunjoSick/filtershosting/actions/runs/37916910885

Source workflow: https://github.com/JunjoSick/filtershosting/actions/runs/37775388864

@JunjoSick
JunjoSick marked this pull request as ready for review October 9, 2026 10:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant