v0.1.0: Auditable memory lifecycle

Initial public release of Agent Memory Lifecycle.

• Validated memory events with provenance, validity intervals and explicit conflicts.
• Deterministic correction, expiry, deletion and sanitized replay semantics.
• Session-scoped local TF-IDF cosine retrieval with auditable selection traces.
• JSON CLI, Python API, MIT license and documented audit retention limitations.
• 18 passing tests with 99% combined statement/branch coverage.
• Real local evaluation: 18/18 policy scenarios pass, compared with 8/18 for the lifecycle-unaware baseline.

Final commit: d185070.

CI: https://github.com/lucioduran/agent-memory-lifecycle/actions/runs/42674645196

Evidence was collected locally on an Intel Xeon W-3223 with Python 3.11.13 and one thread under the shared benchmark lock. The 18 fictional, author-known cases are deterministic policy checks, not production accuracy, security, model-quality or load-test claims. Deletion removes payloads from current exported state but cannot erase external copies or prior traces; audit identifiers and timestamps remain.