IDKMesh

2026-10-13 — Evidence-aware growth selection and quality gates

Owner request

Continue implementing IDKMesh SEO, real traffic and community visibility capability after merging the read-only visibility and contributor-funnel measurement contracts.

Finding

The V6 experiment selector previously ignored the V4 external reference ledger and the V5 funnel-coverage report. It could also recommend changing search snippets from truncated Search Console query evidence, and implicitly coerce some malformed search metrics to zero.

Change

The V6 recommendation-only selector now validates optional finalized Google Search Console query/page evidence, its canonical site prefix, local pagination cap and finite counters. Truncated evidence selects a measurement-quality review before any content/snippet edit.

It also consumes the existing V4 owner-attested reference ledger and V5 coverage-only funnel report. A zero independent-link count means zero reviewed, recorded claims, not zero external backlinks. The V5 report cannot establish cross-source conversions and is forbidden from claiming non-null conversion rates.

The existing read-only weekly observatory now assembles V5 evidence before V6 selection, validates the V4 record and includes an aggregate quality summary in retained artifacts. No new credentials, personal visitor IDs, tracking, issue/PR writes, outreach or merge authority are added.

Verification

Existing and new deterministic selector tests cover old fallback behavior, truncated GSC, malformed/NaN/negative source data, empty versus reviewed authority references, syndication exclusions, invalid funnel attribution, and explicit missing-file failure. Workflow tests assert correct stage ordering and preserved read-only permissions.

No owner Google Search Console property, confidential traffic logs, or independent external reviews were accessed. Exact-head repository CI and standard human review remain required before integration.

Community impact

This prevents spurious SEO changes and makes recommendations about evidence gaps instead of pretending the site has measured conversions or independent citations. Thresholds remain uncalibrated bootstrap assumptions, not an autonomous optimization policy.

Related: issue #664, issue #665, issue #807.

Provenance: AI-assisted source inspection and change; no verified external traffic or independently authenticated review claimed.