fix(site): make policy selection and evidence recovery reliable - #11
Conversation
|
Independent AI landing review: PASS+NOTES. Reviewed head A separate reviewer, who did not author the code, inspected the full diff and independently ran the native verification and committed Chrome suite. Results were 189 tests, 96.58% coverage, all native gates, and 9/9 browser cases. The served source and five protected historical result/evidence files matched their expected bytes. No weakened tests, new suppressions or blocking findings were found. The release coordinator inspected the actual diff, review and evidence. Notes: browser checks remain a separate manual command; the existing trusted checked-in JSON boundary and narrow grid-gap spacing remain. No new hardware benchmark, cross-browser or formal accessibility certification is claimed. This is not an external security audit. Remote CI, merge identity and actual Cloudflare upload remain separate release checks. |
User-visible changes
Not run, without misleading zero-value chart bars.Verification
11e7392789da37041880b8b206ddbfcffa69eecbagainstf40e114c5ed15942f8d797f74df6e95182b2ffa3.55c9eaa49a5ec14b5c89bd7e7113c0aa17244aa5.make verify: 189 tests, 96.58% coverage, Ruff, formatting, strict mypy, deterministic evidence, site checks and distribution checks passed.Scope and limits
Only the static client and its browser check change. Simulator code, schema, scenarios, dependencies and historical numeric evidence do not change. This is not a new hardware benchmark or a formal accessibility certification.
The browser suite is a separate command;
make verifyand existing CI do not invoke it. The existing same-origin checked-in JSON trust boundary and narrow grid-gap spacing remain. Hosted checks and actual Cloudflare artifact verification are separate release gates. No deployment is claimed by this PR description.