Non-technical end users
Hermes had the same wrong-key problem OpenClaw did. One fix solves both.
Campaign a2a-hermes-v0.6.0-r4 FAIL
Infrastructure
Back-filled by scripts/backfill_legacy_runs.sh — historical run predates campaign.meta.json emission.
Run focus
Same SSH key mismatch — identical diagnosis, shared fix lineage
What this campaign tested: Hermes-side SSH path — same key, same secret, same blocker.
What it demonstrated: Confirmed the SSH fingerprint bug is campaign-agnostic (shared secret scope), so fixing it once on the ai2ai-gate repo secret fixes both groups simultaneously. Also confirmed teardown ran clean on Hermes (6 resources destroyed).
AI NHI analysis · Claude Opus 4.7
Same SSH key mismatch — identical diagnosis, shared fix lineage
RED at Wait for SSH after 2 m 38 s.
For three audiences
C-level decision makers
Good economy of failure: the bug was repo-scoped, not agent-group-scoped, so a single secret rotation resolves it for all future dispatches of either group. No 2x fix cost.
Engineers & architects
Since DIGITALOCEAN_SSH_PRIVATE_KEY and DIGITALOCEAN_SSH_KEY_FINGERPRINT are repo-level secrets, both agent_group campaigns consume the same values. Updating the fingerprint secret once makes every future campaign dispatch land on droplets with the correct authorized_keys.
What changes going into the next campaign
Same fingerprint rotation resolves both; redispatch r5 for both groups.