PR #74171: fix(agents): preserve CLI wake-up session metadata

• Repository: openclaw/openclaw
• Author: bitloi
• State: closed | merged: True
• Link: https://github.com/openclaw/openclaw/pull/74171

Description (problem / solution / changelog)

Summary

• Problem: post-restart wake/cron/heartbeat turns for Discord Claude CLI sessions could rebuild CLI prompt identity from synthetic system-event context instead of persisted Discord chat metadata.
• Why it matters: the static prompt hash could drift, causing --resume to be skipped even when the prior CLI session binding was still valid.
• What changed: system-event prompt construction now restores missing prompt-only chat/route metadata from the persisted session entry before building static CLI prompt identity, and CLI exec logs include redacted resume diagnostics.
• What did NOT change: auth, auth-epoch, MCP, and real system-prompt mismatch invalidation remain strict. This does not address the separate group-intro prompt drift tracked in #69118.

Change Type (select all)

• Bug fix
• Feature
• Refactor required for the fix
• Docs
• Security hardening
• Chore/infra

Scope (select all touched areas)

• Gateway / orchestration
• Skills / tool execution
• Auth / tokens
• Memory / storage
• Integrations
• API / contracts
• UI / DX
• CI/CD / infra

Linked Issue/PR

• Closes #74125
• Related #69118
• This PR fixes a bug or regression

Root Cause (if applicable)

• Root cause: wake/heartbeat/cron turns can arrive with Provider=heartbeat|cron-event|exec-event and no live ChatType, while CLI reuse hashes were built from that live context instead of the persisted Discord session entry.
• Missing detection / guardrail: existing tests covered no-reset behavior and static prompt forwarding, but not a post-restart system-event turn rebuilding the same static CLI prompt identity as the prior Discord channel session.

Regression Test Plan (if applicable)

• Coverage level that should have caught this:
  • Unit test
  • Seam / integration test
  • End-to-end test
  • Existing coverage already sufficient
• Target test or file:
  • src/auto-reply/reply/get-reply-run.exec-hint.test.ts
  • src/auto-reply/reply/get-reply-run.media-only.test.ts
  • src/agents/cli-runner.spawn.test.ts
• Scenario the test should lock in: a cron/system-event turn with no live ChatType uses persisted Discord channel metadata for static CLI prompt identity and logs resume diagnostics without exposing raw CLI session IDs.
• Why this is the smallest reliable guardrail: the bug is at prompt identity construction before CLI reuse, so the prepared-reply seam proves the hash input is stable without needing a live Discord or Claude CLI run.
• Existing test that already covers this (if any): none.
• If no new test is added, why not: N/A

User-visible / Behavior Changes

Wake/heartbeat/cron system-event turns targeting a persisted Discord Claude CLI session now keep CLI continuity when auth and MCP identity still match. CLI exec logs also include redacted resume/reuse diagnostics.

Diagram (if applicable)

Before:
post-restart wake -> synthetic cron/heartbeat context -> static prompt hash drifts -> CLI resume skipped -> cold turn

After:
post-restart wake -> persisted Discord prompt metadata restored -> static prompt hash remains stable -> CLI resume can continue

Security Impact (required)

• New permissions/capabilities? (Yes/No) No
• Secrets/tokens handling changed? (Yes/No) No
• New/changed network calls? (Yes/No) No
• Command/tool execution surface changed? (Yes/No) No
• Data access scope changed? (Yes/No) No
• If any Yes, explain risk + mitigation: N/A. Resume diagnostics intentionally log only presence/state and a short one-way session fingerprint, not raw CLI session IDs or argv.

Repro + Verification

Environment

• OS: Linux container
• Runtime/container: Node v24.11.0 via /tmp/node-v24.11.0-linux-x64/bin
• Model/provider: Claude CLI path exercised through unit/seam tests
• Integration/channel (if any): Discord session metadata simulated at the prepared-reply seam
• Relevant config (redacted): persisted Discord channel session entry with stored CLI binding metadata

Steps

1. Start from a persisted Discord channel session entry with CLI session metadata.
2. Simulate a post-restart cron/system-event turn with Provider=cron-event and no live ChatType.
3. Build the prepared reply and inspect the static CLI prompt identity passed toward the runner.

Expected

• The static prompt identity is built from persisted Discord channel metadata.
• CLI diagnostics show trigger/reuse state without leaking raw CLI session IDs.

Actual

• Before this fix, the synthetic system-event context could omit group/channel prompt context and drift the static prompt hash.

Evidence

• Failing test/log before + passing after
• Trace/log snippets
• Screenshot/recording
• Perf numbers (if relevant)

Human Verification (required)

• Verified scenarios:
  • PATH=/tmp/node-v24.11.0-linux-x64/bin:$PATH pnpm test src/agents/cli-session.test.ts src/agents/cli-runner/prepare.test.ts src/agents/cli-runner.spawn.test.ts src/auto-reply/reply/agent-runner-execution.test.ts src/infra/heartbeat-runner.ghost-reminder.test.ts src/gateway/server-cron.test.ts
  • PATH=/tmp/node-v24.11.0-linux-x64/bin:$PATH pnpm test src/auto-reply/reply/get-reply-run.exec-hint.test.ts src/auto-reply/reply/get-reply-run.media-only.test.ts src/agents/cli-runner.spawn.test.ts
  • PATH=/tmp/node-v24.11.0-linux-x64/bin:$PATH pnpm exec oxfmt --check --threads=1 src/auto-reply/reply/session.ts src/auto-reply/reply/get-reply-run.ts src/auto-reply/reply/agent-runner-execution.ts src/agents/cli-runner/prepare.ts src/agents/cli-runner/execute.ts src/agents/cli-session.ts src/infra/heartbeat-runner.ts src/gateway/server-cron.ts CHANGELOG.md src/auto-reply/reply/get-reply-run.exec-hint.test.ts src/auto-reply/reply/get-reply-run.media-only.test.ts src/agents/cli-runner.spawn.test.ts
  • PATH=/tmp/node-v24.11.0-linux-x64/bin:$PATH pnpm check:changed
• Edge cases checked:
  • missing system-event ChatType
  • synthetic heartbeat/cron-event provider labels
  • persisted Discord channel route metadata
  • normal user turns remain on live metadata
  • explicit system-event metadata is not overwritten
  • raw CLI session IDs are not logged
• What you did not verify:
  • a live Discord + Claude CLI gateway restart repro

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/No) Yes
• Config/env changes? (Yes/No) No
• Migration needed? (Yes/No) No
• If yes, exact upgrade steps: N/A

Risks and Mitigations

• Risk: restoring persisted prompt metadata too broadly could mask a real auth/MCP/system-prompt mismatch.
• Mitigation: this only changes prompt context reconstruction; CLI reuse invalidation for auth profile, auth epoch, MCP, and durable prompt hash mismatch remains strict.

Changed files

• src/agents/cli-runner.spawn.test.ts (modified, +26/-1)
• src/agents/cli-runner/execute.ts (modified, +51/-1)
• src/auto-reply/reply/get-reply-run.exec-hint.test.ts (modified, +95/-0)
• src/auto-reply/reply/get-reply-run.media-only.test.ts (modified, +58/-0)
• src/auto-reply/reply/get-reply-run.ts (modified, +129/-16)