chore(deps): align knowledge with Eval 0.144.1

[drewstone]

Problem

Published Knowledge 7.0.8 pins Agent Eval 0.143.0. Runtime 0.128 requires Eval 0.144.1, so a combined packed install resolves the older campaign declarations and cannot compile Runtime's official-optimizer callback type.

Change

• publish Knowledge 7.0.9 on exact Eval 0.144.1
• adopt Eval's caller-owned SkillOpt model callback in the Knowledge contract test
• record explicit observed/uncaptured cost origin in optimization fixtures
• compile-import ExternalOptimizerModelCall in the stack contract

Direct proof

• pnpm run lint
• pnpm run typecheck
• pnpm run test — 545 passed, 12 skipped
• pnpm run build
• pnpm run verify:package — one installed Eval 0.144.1, Core 0.4.33, Interface 0.43.0
• pnpm run verify:official-optimizers

[tangletools]

✅ Auto-approved drewstone PR — 9fc3c6d7

This PR was opened by the trusted drewstone account.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

_{tangletools · auto-approval · reason: drewstone_author · 2026-08-04T02:55:30Z}

[tangletools]

✅ Auto-approved drewstone PR — ad03246a

This PR was opened by the trusted drewstone account.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

_{tangletools · auto-approval · reason: drewstone_author · 2026-08-04T03:08:28Z}