Summary

• Problem: Internal hooks and before_reset plugin hook only fire for manual /new and /reset commands, not for daily (4AM) or idle-timeout session resets. This means session-memory never saves on daily/idle resets.
• Why it matters: 4 community issues report lost session memory on daily/idle resets (Feature: session:end internal hook event #10142, session-memory hook should support session:end events #31266, Session reset hooks: idle/daily reset do not trigger session-memory save #50891, Daily reset doesn't trigger session-memory hook flush #43524)
• What changed: Wire both hook types into the lazy staleness path in initSessionState(), update session-memory registration and filter
• What did NOT change (scope boundary): No new hook event names, no changes to plugin SDK API, no extraction/refactor — only export added to existing functions

Change Type (select all)

• Bug fix
• Feature

Scope (select all touched areas)

• Memory / storage

Linked Issue/PR

• Closes Feature: session:end internal hook event #10142
• Closes session-memory hook should support session:end events #31266
• Closes Session reset hooks: idle/daily reset do not trigger session-memory save #50891
• Closes Daily reset doesn't trigger session-memory hook flush #43524
• This PR fixes a bug or regression

Root Cause (if applicable)

• Root cause: initSessionState() fires session_end/session_start plugin hooks for lazy resets but never fires triggerInternalHook or hookRunner.runBeforeReset. These only fire via emitResetCommandHooks() in commands-core.ts, which is only called for manual /new and /reset commands.
• Missing detection / guardrail: No test asserted that internal hooks fire for lazy resets
• Contributing context (if known): The lazy evaluation design means sessions aren't proactively killed — staleness is checked on next message arrival, a code path separate from manual commands

Regression Test Plan (if applicable)

• Coverage level that should have caught this:
  • Unit test
• Target test or file: src/auto-reply/reply/session.stale-hooks.test.ts (new), src/hooks/bundled/session-memory/handler.test.ts (extended)
• Scenario the test should lock in: Internal hook fires with action: "daily" / "idle" when session is stale; before_reset plugin hook fires for lazy resets; no double-fire on manual reset
• Why this is the smallest reliable guardrail: Mocks triggerInternalHook and hookRunner to assert dispatch without side-effects
• Existing test that already covers this (if any): session-hooks-context.test.ts covers session_end/session_start but not internal hooks or before_reset

User-visible / Behavior Changes

session-memory now saves session summaries on daily (4AM) and idle-timeout resets, not just manual /new and /reset. Plugin before_reset hook now fires for lazy resets, giving plugins pre-archive transcript access.

Diagram (if applicable)

Before:
[lazy reset] -> session_end + session_start hooks only

After:
[lazy reset] -> internal hook (command:daily/idle) + before_reset + session_end + session_start

Security Impact (required)

• New permissions/capabilities? No
• Secrets/tokens handling changed? No
• New/changed network calls? No
• Command/tool execution surface changed? No
• Data access scope changed? No

Repro + Verification

Environment

• OS: Any
• Runtime/container: Node.js 24
• Integration/channel (if any): Any channel with daily or idle reset configured

Steps

1. Configure a session with daily reset (default 4AM)
2. Have an active session with messages before 4AM
3. Send a message after 4AM (triggering lazy staleness detection)

Expected

• session-memory saves a summary of the previous session
• before_reset plugin hook fires with reason: "daily" and transcript messages

Actual (before this PR)

• session-memory does not fire — no summary saved
• before_reset plugin hook does not fire

Evidence

• Failing test/log before + passing after

Human Verification (required)

• Verified scenarios: Daily reset with stale session triggers session-memory save; idle reset triggers save; manual /new does not double-fire
• Edge cases checked: First-ever session (no previous entry), system events (heartbeat), fresh session
• What you did not verify: End-to-end with real LLM slug generation (mocked in tests)

Review Conversations

• I replied to or resolved every bot review conversation I addressed in this PR.
• I left unresolved only the conversations that still need reviewer or maintainer judgment.

Compatibility / Migration

• Backward compatible? Yes
• Config/env changes? No
• Migration needed? No

Risks and Mitigations

• Risk: before_reset receives the post-archive sessionFile path (after renameSync), not the original path. loadBeforeResetTranscript handles ENOENT by scanning for the latest .reset.* archived sibling.
  • Mitigation: This matches the manual reset path behavior. The stable post-archive path is intentional — it avoids racing transcript reads against the archive rename.

AI Disclosure

• Mark as AI-assisted in the PR title or description
• Note the degree of testing: fully tested — 10 unit tests (7 new + 3 extended) + manual daily/idle reset verification
• Confirm: I understand what this code does and have verified the hook dispatch logic, guard conditions, and fire-and-forget patterns

🤖 AI-assisted with Claude Code

cc @vincentkoc (hooks/plugins CODEOWNER)