From the 2026-07-24 fleet spec (H2). Depends on the repo registry (H1) and the runtime probe (NimbleCoOrg/hermes-agent-mt#117).
Problem
The 2026-07-24 NimbleCoAI→NimbleCoOrg transfer turned into a day-long outage because "a repo moved" had no operational shape — each consequence (dead grants, stale remotes, pinned credentials) was discovered separately, in production, by the agent.
Proposal
A single HSM operator flow for a transfer:
- Verify the GitHub App installation covers the new owner (and the repo, if the installation is on selected-repos).
- Update the registry row (H1), appending to
moved_from[].
- Trigger fleet re-provision via the existing HSM API (runtime R1/R2 do the in-container healing).
- Probe fleet-wide (runtime R3) and show green/red per agent.
Turns this week's day-long fight into a two-minute checklist.
Non-blocker follow-up (spec H3)
The flow should file a docs/skills cleanup task (grep for {moved_from}/ across skills + memory) rather than gate on it — ops must survive on redirect-following (hermes-agent-mt#116) while docs lag.
From the 2026-07-24 fleet spec (H2). Depends on the repo registry (H1) and the runtime probe (NimbleCoOrg/hermes-agent-mt#117).
Problem
The 2026-07-24 NimbleCoAI→NimbleCoOrg transfer turned into a day-long outage because "a repo moved" had no operational shape — each consequence (dead grants, stale remotes, pinned credentials) was discovered separately, in production, by the agent.
Proposal
A single HSM operator flow for a transfer:
moved_from[].Turns this week's day-long fight into a two-minute checklist.
Non-blocker follow-up (spec H3)
The flow should file a docs/skills cleanup task (grep for
{moved_from}/across skills + memory) rather than gate on it — ops must survive on redirect-following (hermes-agent-mt#116) while docs lag.