S6: Add token_estimate field to capability metadata schema

[normenmueller]

Parent Epic

#37

Scope

Add a token_estimate field to the central capability metadata schema. This field holds a deterministically computed token estimate per capability (derived from word count of the .md file, no LLM required). The field enables downstream context budget checks during capability composition.

Acceptance Criteria

AC-21 (Ubiquitous): The central schema definition file shall declare an optional field token_estimate of type integer.

AC-22 (Ubiquitous): A CI script shall compute token_estimate for each capability by applying a deterministic word-count-based formula to the capability .md file and shall verify that the value in .meta.yaml matches the computed value within a defined tolerance.

AC-23 (Ubiquitous): All existing .meta.yaml files shall be populated with a computed token_estimate value.

AC-24 (Unwanted): If token_estimate is present but deviates from the computed value beyond the defined tolerance, the CI validation shall report it as an error.

Dependencies

• S1: Define central capability metadata schema #38 (S1: schema file must exist)

Notes

• Budget ceiling and tiered truncation strategy are out of scope for this Story — they depend on curate pipeline design and are parked until curate is conceptually defined.
• Origin: adm/pbl/ai4x-context-budget-model.md (partially absorbed; budget/truncation portions remain parked).