feat(cohort)!: adopt unified capture foundations

[drewstone]

Change

Release Knowledge 18 against the compatible evidence cohort:

• Eval >=0.201.0 <0.204.0, with development pinned to 0.203.0.
• Interface 2.15, Zod 4.6.5, and public TCloud 0.8 / Sandbox 0.58.4.
• Node >=22.12, matching the maintained TCloud runtime floor.

This is a major release because the accepted peer window and Node floor narrow.
Knowledge operations and stored record formats are unchanged.

Verification

• Normal install from public packages; no contract override.
• Full source suite: 915 passed, 12 skipped.
• Source and contract types, lint, build, and API surface passed.
• Packed consumer and actual official GEPA/SkillOpt integration passed separately with Eval 0.201.0, 0.202.0, and 0.203.0.
• Each optimizer proof used its matching published Python RPC package.

GTR logs are retained under /tmp/knowledge-18-*-20260929.log.
No scientific pursuit or model research was launched by these engineering checks.

Delivery

The exact committed source is independently reviewed before merge.
The maintained tag publisher and installed public-package proof remain the publication checks.

[tangletools]

✅ Auto-approved PR — dc1794f9

Blanket team auto-approval is intentional. The merge gates are CI and codex-p1.
No automated review runs on this PR. This approval rests on the rule above alone.

_{tangletools · auto-approval · reason: blanket_auto_approve · 2026-09-30T05:34:32Z}

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.