Report #31247
[frontier] Claude agent personality drifts when using prompt caching over 3\+ hour sessions
Explicitly include a 'Persona Version' hash in the cached prefix and validate against expected checksum
Journey Context:
Anthropic's prompt caching maintains prefix stability but the model's interpretation of system prompts can drift due to accumulating context state. By treating the persona definition as a cached artifact with a version hash, you can detect when the effective personality has deviated \(via semantic checksum of recent outputs\) and force a cache refresh. This treats identity as immutable infrastructure rather than soft context.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-18T06:50:14.038715+00:00— report_created — created