Skip to content

publication: retire the reader-test approval gate (owner decision 2026-11-13) - #46

Merged
Cubits11 merged 1 commit into
mainfrom
claude/retire-reader-test-2026-11-13
Nov 13, 2026
Merged

Cubits11 merged 1 commit into
mainfrom
claude/retire-reader-test-2026-11-13

Conversation

@Cubits11

Copy link
Copy Markdown
Owner

The deployment gate required five unfamiliar human readers, four passing each of three tasks, before it would publish. Zero participant records were collected. The last deployment that published is 50b2898 (2026-10-17, run 39066600697). Eleven pushes to main since stopped at the deployment step; from 7764a42 onward this rule was the only blocker (runs 41591829298, 41594883017, 41609925565).

By owner decision on 2026-11-13 the rule is removed and the updated site ships without it. No replacement approval step is added.

Changed

  • scripts/publication_gate.py checks the title attestation and the phone-film receipts only. The reader check, --bind and --template are removed.
  • tests/test_publication_gate.py drops the reader cases. One new case holds that reader records are not required; one holds that a blocked gate exits 2.
  • publication/readers.json, an empty study record, is deleted.
  • publication/readiness.json records the decision under reader_test: status, date, the rule as it stood, when it was in force, what was removed and kept, and non-claims.
  • publication/reader-test-kit.md is marked RETIRED and kept as the record of what the rule asked.

Not changed

  • The title and phone-film checks, the claim registry, reproduction and deployed-site smoke jobs still gate every deploy.
  • The deploy step name in .github/workflows/verify.yml and one docstring line in scripts/verification_manifest.py still mention reader evidence. Both are trust-root files under security/evidence-policy.json; this PR does not touch them, so the evidence boundary stays green. Rewording them is a separate migration.
  • The film cold-viewer rule for feed releases (campaigns.yaml, distribution/launch-units.yaml) is a different rule and stays.

Non-claims

Shipping without the test does not establish that unfamiliar readers understand the site. REL-001 stays untested.

Checked locally

  • python scripts/verification_manifest.py: 90 unique checks, exit 0.
  • python scripts/publication_gate.py: READY, no blockers.
  • python scripts/evidence_guard.py --base <origin/main> --candidate <head>: PASS.
  • python scripts/verify_attribution.py: all Cubits11, no machine trailers.

Merging deploys origin/main as of c1642d6 plus this change. That replaces the live 50b2898 build.

…on, 2026-11-13

The deployment gate required five unfamiliar human readers, four passing each
of three tasks, before it would publish. Zero participant records were
collected. The last deployment that published is 50b2898 on 2026-10-17; eleven
pushes to main since stopped at this gate, and from 7764a42 it was the only
blocker.

The owner removes the rule and ships the updated site without it. No
replacement approval step is added. publication_gate.py now checks the title
attestation and the phone-film receipts only; --bind and --template are gone
with the study they served. readers.json, an empty study record, is deleted.
readiness.json records the decision under reader_test, with the dates it was
in force and what it does not license. The kit is marked retired and kept as
the record of what the rule asked.

Not changed: the deployment step name in verify.yml and one docstring line in
verification_manifest.py still mention reader evidence. Both are trust-root
files and change only through a separate migration. The film cold-viewer rule
for feed releases is a separate rule and stays.
@Cubits11
Cubits11 merged commit 3b6d994 into main Nov 13, 2026
6 checks passed
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