LLM Anchor Section: Add Canonical Summary to llms-full.txt

[johnx25bd]

Priority: Medium

Description

If the purpose of llms-full.txt is to help models answer accurately, the current structure causes predictable failure modes:

1. Repetition without hierarchy: Same concepts appear in multiple places; models blend them
2. Conflicting definitions: Schema shapes differ; models hallucinate hybrids
3. Too many ... placeholders: In code and UIDs; LLMs treat these as literal patterns
4. Links as authority: References GitHub issues/specs, but models can't follow links

Fix Direction

Add an "Authoritative Canonical Summary" section (1-2 pages) at the top of llms-full.txt defining:

• Entities: Location Attestation, Policy Attestation, InputRef, Delegated Attestation
• Canonical schemas: Or explicit per-chain variants
• Canonical SDK method signatures and return types
• Key terminology and precise definitions

Then append the detailed reference dump. The dump is useful, but it needs an anchor.