Intercept index vNext: structured worked / failed feedback

[aural-psynapse]

What

Add structured feedback fields to intercept records that record what worked, what failed, and an optional failure reason in a queryable shape.

Why

Clear outcome signals improve triage, trend analysis, and any future automation built on top of the intercept index. Today, success / failure is encoded inconsistently across paths (free-text error strings, missing fields, layered statuses).

Scope

• Define a small, stable feedback schema (status + reason / category).
• Persist feedback alongside each intercept row in a queryable form.
• Surface feedback through any read APIs that already expose intercept metadata.

Suggested feedback schema (v1)

• status — success | failure | partial.
• category — e.g. network, policy, verification_mismatch, timeout, runtime.
• reason_code — short, stable, machine-friendly code.
• reason_detail — human-readable message.
• scope — run | action | claim.

Design constraints

• Keep the taxonomy small and stable at first; resist category explosion.
• Separate machine fields (category, reason_code) from human text (reason_detail).
• Allow partial failure to be represented per claim / action without losing the run-level outcome.

Done when

• Schema and enums defined in the SDK.
• Existing failure paths mapped into the taxonomy.
• Read API exposes feedback cleanly.
• Unit tests cover raw-error -> structured-feedback mapping.
• Integration tests cover at least one success and several failure categories.
• Backward-compatibility tests cover records lacking feedback.

Depends on

• Intercept index vNext: capture intent and execution context #4 (intercept index vNext: intent + execution context) — share schema-evolution approach.