$ the-wire · showcase
Runtime auth snapshots preserved, AWS SSH ingress churn fixed
By RepoJournal · Filed · About OpenClaw
OpenClaw preserves runtime-only external OAuth profiles across saves, and Crabbox stops pruning active AWS SSH leases during refreshes.
OpenClaw now keeps external CLI OAuth profiles available in-process after a save operation, instead of dropping them into disk persistence. The change updates save/merge/read paths to filter only what is written to `auth-profiles.json` and `auth-state.json`, preserving runtime access for profiles like `anthropic:claude-cli` while excluding inherited main credentials [1]. This addresses a reported credential drop where the old behavior wrote the disk-filtered store into an existing runtime snapshot [2].
The WhatsApp gating layer now logs a single warning per `accountId:conversationId` when a group inbound message is dropped because the group is not registered in `channels.whatsapp.groups`. Previously this only appeared in verbose mode; operators will now see the exact config path to edit without enabling verbose logging [3].
In Crabbox, AWS SSH ingress churn is fixed by moving source reconciliation behind provider hooks. Previously, the AWS client treated the CIDR list passed to security-group refresh as the desired managed SSH ingress set, pruning Crabbox-owned rules outside that set, so another create or heartbeat from a different source IP could remove an active lease [4]. AWS managed leases share the `crabbox-runners` security group, and the fix preserves active lease source CIDRs [5]. Related refactors route provider-owned behavior through capabilities for both the worker [6] and CLI [7], and stale coordinator leases are retried once after an SSH drop before a command starts [8].
Action items
- → Update OpenClaw to include the runtime auth snapshot fix if you use external CLI OAuth profiles openclaw/openclaw [plan]
- → Upgrade Crabbox worker to prevent AWS SSH ingress pruning of active leases openclaw/crabbox [immediate]
- → Watch for the WhatsApp warning log on missing channels.whatsapp.groups entries openclaw/openclaw [monitor]
References
- [1] Preserve runtime external auth snapshots (#85558) ↗ openclaw/openclaw
- [2] Preserve runtime external auth snapshots ↗ openclaw/openclaw
- [3] fix(whatsapp): warn once when group inbound dropped for missing channels.whatsapp.groups entry ↗ openclaw/openclaw
- [4] fix(worker): preserve active AWS SSH ingress ↗ openclaw/crabbox
- [5] fix(worker): preserve active AWS SSH ingress ↗ openclaw/crabbox
- [6] refactor(worker): route provider behavior through capabilities ↗ openclaw/crabbox
- [7] refactor: route CLI provider behavior through capabilities ↗ openclaw/crabbox
- [8] fix: retry stale coordinator lease before command ↗ openclaw/crabbox