Clear diagnostics when an agent's Conderis MCP connection is missing or unauthorized #371
Description
Problem
In about half of the Codex exec runs of one session the conderis MCP was either not loaded or answered 401 (memory_recall on an old project-memory endpoint). The agents then guessed: tried other endpoints or browser logins, or silently skipped claiming/completing, leaving task state inconsistent.
Proposal
- A documented, copy-paste recovery for Codex/Claude (e.g.
codex mcp login conderis, config key to check) in the guide and in the 401 response body. - A lightweight health/whoami tool so an agent can check at start whether it is connected and as whom.
- Guide text: if Conderis is unreachable, do not try other endpoints; report it so the coordinator can update the tasks.
Evidence
Hyperion Drift 2026-09-27: tasks #129, #130, #131, #146, #147, #153 had to be claimed/completed by the coordinating Claude session because the Codex runs had no Conderis access.