Summary

Severity: Medium-High
Area: Sessions
Affected files: agentkit/core/agent.py:176-180, agentkit/memory/redis_store.py:61-70
Suggested branch: fix/session-history-load-failure
Suggested PR title: fix(agent): stop silently discarding session history

This is finding AUDIT-10 of 18 from a consolidated audit of agentkit/. It is deliberately self-contained so it can be fixed on its own branch with its own PR.

Related findings: Related: AUDIT-04, AUDIT-11.

AUDIT-10 — Medium-High — Session-history load failure is silent and destructive

File — agentkit/core/agent.py:176-180

            messages: list[Message] = []
            if session_id and self.memory is not None and hasattr(self.memory, "get_messages"):
                with contextlib.suppress(Exception):
                    loaded = await self.memory.get_messages(session_id)
                    if loaded:
                        messages = list(loaded)

Why this is a bug

If the session history cannot be read (Redis down, corrupt JSON, a Message that fails validation — note RedisMemoryStore.get_messages returns [] on any deserialisation error, redis_store.py:61-70), the agent silently behaves as if it were a brand-new conversation. The subsequent save_messages then overwrites the existing history with the truncated new one. The result: multi-turn context is permanently lost, with no log line and no error in the response.

Pass cases

• A failure to load session history is logged at ERROR/WARNING with the session_id.
• A load failure does not silently replace the stored history: either the run fails explicitly, or the existing history is preserved (no destructive overwrite).
• A corrupt/unreadable history entry (e.g. invalid Message payload) is surfaced rather than reduced to an empty list.
• Regression test: force get_messages to raise, assert (a) a log record exists, (b) the previously stored messages are still present afterwards.

Definition of done (per AGENTS.md)

• New/updated unit and integration tests covering the pass cases above, all passing (pytest).
• ruff check ., ruff format --check . and mypy agentkit clean.
• Docstrings updated where behaviour changed; docs/PROGRESS.md updated (status, PR link, commit count, notes).
• A new ADR in docs/DECISIONS.md if an architectural decision was made.
• Branch pushed and PR opened (never commit directly to main).

Baseline at the time of writing: pytest is green on main (all tests pass; tests/real/* skipped unless AGENTKIT_REAL_INFRA=1).