26-discipline validation — actual local results

Release: 2026.09.09-d26. These are artifact/tool validation results, not production-site acceptance or an independent external audit.

Executed results

Check family	Result	Boundary
Utility unit tests	84 passed: 24 baseline + 60 new	Local synthetic fixtures; no real website/account
CLI smoke tests	6 passed	Actual local invocation, overwrite rejection, template blocking and malformed-input handling
Browser interaction checks	23 passed	Generated HTML rendered in headless Chromium 144.0.7559.96 through set_content
Role coverage	26 exact IDs with charter, prompt, brief, sources and handoffs	Presence/linkage, not 26 executed specialist audits
Preserved originals	146 hashes verified	Original bytes preserved, not all legacy facts validated
Local Markdown file targets	3892 checked; zero missing/outside targets	No Markdown-fragment or external-URL completeness claim
Markdown files audited	267	Active Library, Publishing, Templates, Tools, Research, Reports and root
JSON files parsed	60	Historical originals excluded; parsing is not semantic truth

The unit tests exercise routing for forms, DNS/tailnet, outages, AI citations, mail policy, explicit overrides and unknown tasks; bounded context packs; gate rejection of missing/blocked/failing checks, absent or changed files, path traversal, symlink escapes, contract/hash/build/run mismatches, reused identities, weak independence declarations, unauthorized exceptions, duplicate IDs/keys, future/naive timestamps, placeholders and publication-review gaps. Synthetic structurally valid fixtures are explicitly not client evidence.

Browser review

Verified report load, document counts, all 26 role charters, role/full-text search, current DMARC references, internal brief navigation, history restoration, prompt companion paths, legacy boundaries, six draft entries, sanitized article DOM, keyboard-scrollable table regions, no horizontal overflow at a simulated 390-pixel viewport and no automatic external requests or JavaScript runtime errors. Desktop and narrow screenshots were visually inspected for legibility and layout.

Local HTTP navigation was blocked by the environment with ERR_BLOCKED_BY_ADMINISTRATOR. The generated HTML was therefore rendered using Chromium set_content, not through a hosted live site or a successfully navigated local HTTP origin. Actual Safari, Firefox, branded Chrome, physical iPhone/Android and assistive-technology combinations were not tested. These checks do not establish full WCAG 2.2 AA conformance.

Gate truth boundary

The evidence checker can detect inconsistent records and verify local bytes. It cannot authenticate an approval, a human identity, reviewer independence or the truth of a captured statement. Every structural pass retains authorization_granted: false, external_truth_verified: false and identity_or_independence_authenticated: false. Template acceptance records intentionally fail; no production acceptance record has been fabricated.

Reports and reproducibility

• Unit output
• CLI smoke results
• Browser results and limits
• Library integrity result
• Desktop screenshot and simulated narrow screenshot

Run python3 -m unittest discover -s Tools/tests -v and python3 Tools/library_audit.py . from the extracted library root. The original legacy reports remain historical baseline snapshots. No external site, Bubbles service, DNS record, mailbox, analytics account, repository or persistent Library folder was changed.