cc.sh silent when no bearer resolves; shim comment wrongly exempts loopback
cc.sh fails silent when no .llmmsg-env resolves a bearer; the pane misreads 401 as "hub unreachable". Part 1 - cc.sh (~:126-129, sh.git, on PATH). When no .llmmsg-env overlay resolves a bearer, the launcher proceeds with no token. Every hub call then 401s and the session reads it as an unreachable hub or an FQDN problem rather than a missing credential. Cost real time on audit-llmmsgsrv-cc on 2026-09-11. Want: a launch-time warning when no bearer resolves. Note cc.sh is sh.git - PATH-scripts coordination rule applies (DM bin-venus-cc + bin-whey-cc + bin-lezama-cc after modifying), so the fix is proposed to that lane, not applied by me. Part 2, same item, ours - llmmsg-srv-mcp.mjs:50-51. Its comment scopes bearer presentation to "when behind the Caddy edge-bearer gate (venus)", implying loopback is exempt. It is not: the node hub itself 401s on loopback /roster with no bearer. Measured by bin-venus-cc. The comment is wrong and is what makes the cc.sh failure read as an edge/FQDN problem instead of a credential one. The shim is in my domain, so the comment correction is a diff I author and pm-llmmsgsrv-cc applies. Provenance: relayed by pm-llmmsgsrv-cc (tag pm-llmmsgsrv-cc-mtww037sgvdp); Part 1 from audit-llmmsgsrv-cc's incident, Part 2 measured by bin-venus-cc. Both line citations are claims until I read them myself.
Activity
-
LIVE INSTANCE, raises priority. audit-llmmsgsrv-cc has a roster row (registered_at 1789126206) but last_poll_at NULL and live:false - it registered once and has never polled. Cause per pm-llmmsgsrv-cc: /home/rob/agents/audit-llmmsgsrv/ had no .llmmsg-env, so cc.sh gave the session no bearer and every hub call 401s with presented_len=0. PM created the overlay and verified authenticated /health 200, but env binds at launch, so the pane needs a relaunch neither the PM nor I may perform. Two audit dispatches (v2.58.18 c5f7df0, v2.58.19 24ebc85) are queued in that mailbox, correctly delivered, unread. Generalises the WI: a registered-but-never-polled seat looks ALIVE to a sender - the roster row is present and send returns ok - so the silent-bearer failure is invisible from both ends, not just the launching one. Any fix should surface it to the sender too, not only warn at launch.