trace-repair-suspect

Trace repair suspect proof

This offline example turns a signed trace packet into a repair target. It creates three current trace records, rejects a stale prior packet, ranks the failed tool action as the likely suspect, and signs a diagnostic outcome that cites both the failure and the suspect through informed_by.

Run it with:

pnpm --filter @atrib/integration trace-repair-suspect

The example proves:

• A current trace packet can be verified with signer, context, body-commitment, freshness, and log-inclusion checks before any follow-up work cites it.
• A stale prior packet is rejected before it can steer the repair.
• Suspect ranking stays derived analyzer output. The base records and graph edges remain structural.
• The diagnostic outcome is a signed atrib record with informed_by links to the failure record and the top suspect.

This is not an AgentTrace implementation and it is not a Tracebase memory runtime. It is the atrib-side adapter shape: trace tools can keep their local UI, ranking, and reports, while atrib supplies signed records and verifier-ready handoff material.