Skip to content

中文版本

A2 | Event Consumer Visibility

Claim

Historical storage containing projected fields does not prove that every query and outbound consumer applies the same least-visibility contract. In the tested V2.0.4 Activity path, an ordinary query returned a unique marker that existed only inside payload.raw. That is an old-baseline result; the delivered V2.1.2 behavior is documented below.

Historical data profile

DatasetRowsWith payload.rawShare
Runtime2,7431,47453.7%
Analytics17,69716,82895.1%
Total20,44018,30289.5%

A later Analytics subset from August 10 and 12 contained 681 rows with payload.raw=0. This is a sample observation, not a global field-allowlist guarantee.

Re-runnable public attachments

The Reader/check below reads frozen JSON and verifies the statistics and probe record. It does not execute the private Runtime. The current field in the output refers to the original experiment baseline, not the version current when this page is read.

text
node A2-event-consumer-check.mjs

Expected output:

json
{"evidence_id":"RBE-20260828-A2","rows":20440,"rows_with_payload_raw":18302,"raw_percent":89.5,"runtime_raw_percent":53.7,"analytics_raw_percent":95.1,"current_query_returned_raw_marker":true,"status":"PASS"}

Boundary

The historical overall share is not a current V2.0.4 share. Raw fields appearing in a query result do not by themselves prove unauthorized access occurred. The public fixture does not cover Panel, Webhook, LAN or every diagnostic entry point.

2026-08-30: V2.1.2 delivered update

A pre-fix rerun on V2.1.1 36e5c83b confirmed that the raw marker could still be returned. V2.1.2 introduced server-defined web_panel / activity_api / analytics consumer policies and reconstructs new objects from recursive field allowlists. Unknown events do not pass unknown payloads through, while raw internal events remain available to authorized internal paths.

The first Shell targeted regression was 19 pass / 1 fail because semantic warning count dropped from 1 to 0 after an allowed diagnostic signal was over-pruned. After deriving and projecting only ok/code/summary_blocked_reason/projection_status, the targeted rerun reached 20/20 without restoring raw pass-through.

Independent QA B1 on candidate 64f633ac observed one event, raw-marker count 0 and raw_present=false; event_type, task_id, session_id and projected_summary remained. B1 is an internal acceptance identifier, not this public A2 pack number.

Sources for the newer claims and review visibility

ClaimFirst-party sourceVisibility
Pre-fix marker visibility and allowlist implementationRUNTIME-BOUNDARY-20260830-001, implementation 3302ca61Restricted mother-repository material
Over-pruning regression and repair, 19/20 → 20/20Failures and targeted-final records in the same packRestricted mother-repository material
B1 independent validationRUNTIME-BOUNDARY-QA-20260830-001: independent-qa recordsRestricted mother-repository material
V2.1.2 release and final regressionV2.1.2-R3, V2.1.2-PUBLICATION-20260830-001; Runtime 1842 pass / 0 fail / 1 skip, Shell 1037 pass / 0 failFirst-party release record, 2026-08-30

The CodeFlowMu V2.1.2 Public Release Summary consolidates recursive allowlisting, safe projection, B1 independent QA, compatibility, release gates and residual risk. Complete source and raw release material remain restricted first-party evidence and are not public reading links; the historical JSON is not presented as a public fixture that can fully rerun the new private product.

The three current consumer projections are not a universal sensitive-content detector, complete subscriber governance or a real-network authorization audit. Open Edition was not released and no online instance is implied to have switched. Existing dependency warnings, the symlink-permission skip and uncovered real LAN/Gateway scope remain explicit. The historical 681-row raw=0 sample also cannot substitute for the new implementation tests.

Last updated: