(agent): Machine-readable engineering standards (repo codex)

## Component

Agent (Python runtime)

## Describe the feature

Support a versioned, machine-readable **engineering standards** artifact per repo (e.g. `.abca/standards.yaml` or markdown with schema) that automated review and agents map findings to—so review output is objective (“violates standard X”) not subjective (“maybe improve this”).

## Use case

AI code review is ignored when it sounds like opinion. We should codify standards, link findings to rule IDs, and track when humans override (**break-glass**). ABCA has Cedar and `pr-review` but no first-class **standards document** contract.

## Proposed solution

1. Define schema for standards rules: `id`, `severity`, `description`, `applies_to` globs, `enforcement` (advisory | gate).
2. Hydration injects standards summary; `pr-review` prompts require citing `rule_id`.
3. Optional CI check: standards file lint + drift vs. Cedar policies.
4. Integrate with **feature-pre-human-automated-review-gate** and **feature-automated-review-break-glass-tracking**.

## Other information

- Complements **feature-service-catalog-context-hydration** (service metadata vs. coding standards).
- Feeds roadmap **Review feedback memory loop**.

## Acknowledgements

- [ ] I may be able to implement this feature
- [ ] This might be a breaking change

## Acceptance criteria

- [ ] Schema documented; example committed under `docs/` or sample repo.
- [ ] `pr-review` structured output includes `rule_id` when standards configured.
- [ ] Blueprint references standards path.
- [ ] Tests for hydration and review citation.

Comienza por localizar Agent Python runtime, hydration flow, pr-review integration, Blueprint y las Cedar policy references. Revisa cómo intercambian actualmente review context estos componentes y, después, utiliza los criterios de aceptación para definir el standards contract, la documentación o el ejemplo bajo docs/, las structured rule citations y las pruebas de hydration y review como comprobaciones de finalización.

Escrito por el modelo de indexación a partir del texto del issue.