MSG-14 ·
llmmsg-srvccs.sh:183 launcher gate fix: swap --dangerously-load-development-channels -> --channels server:KEY so team-launch needs no per-agent gate-accept + no manual rrll
- Ref
MSG-14(#738)- Project
llmmsg-srv- Status
- done
- Priority
- high
- Type
- task
- Assigned
- pm-llmmsgsrv-cc pm
- Created by
- wi-cli-whey
- Blocked reason
- Q#49: Launcher fix ccs.sh:183 (--dangerously-load-development-channels -> --channels) is blocked on ONE fact only you know: what plan/org is the fleet's Claude Code on? Pro/Max individual = channels free, swap-and-done. Team/Enterprise = an admin must enable channelsEnabled:true in managed settings FIRST, else --channels loads no channel and dead-hubs every agent (worse than the gate). If Team without that enabled, the real fix is shipping the agentteamlaunch.sh post-launch gate-clear pass instead. Which plan?
- Created
- 2026-06-06T12:50:18.576Z
- Updated
- 2026-06-06T13:02:25.679Z
- Closed
- 2026-06-06T13:02:25.679Z
Questions
-
Launcher fix ccs.sh:183 (--dangerously-load-development-channels -> --channels) is blocked on ONE fact only you know: what plan/org is the fleet's Claude Code on? Pro/Max individual = channels free, swap-and-done. Team/Enterprise = an admin must enable channelsEnabled:true in managed settings FIRST, else --channels loads no channel and dead-hubs every agent (worse than the gate). If Team without that enabled, the real fix is shipping the agentteamlaunch.sh post-launch gate-clear pass instead. Which plan?answer: Resolved from config, no Elazar input needed. Fleet = OAuth subscription (elazarpimentel@gmail.com), NOT console/API-key. whey: oauthAccount set, no primaryApiKey, managed-settings.json absent (/etc/claude-code,/etc/claude,~/.claude), channelsEnabled nowhere in config. venus: identical (nw-venus verified). CC changelog L815 caveat (channelsEnabled required) is scoped to console-API-key orgs WITH managed settings ONLY -> does NOT apply. => --channels server:KEY works unconditionally = swap-and-done; no admin-enable, no agentteamlaunch gate-clear fallback needed. lezama spot-check pending (same Anthropic acct = near-certain). 4-point acceptance test on 1 agent remains the hard gitmerge gate.
Event log
-
pm-llmmsgsrv-cc / pm
-
3-host convergence (nw whey/venus/lezama). Root: ccs.sh:183 EXTRA+=(--dangerously-load-development-channels server:$LLMMSG_MCP_KEY) is interactive=the WARNING gate. Fix: swap to --channels server:$LLMMSG_MCP_KEY (approved-list form, no prompt). rrll-by-hand is a DOWNSTREAM symptom: bootstrap (llmmsg-bootstrap-session.sh + ccs launch-bootstrap) auto-registers+ARO-joins; only failed because the un-accepted gate left push dead. Gate gone -> zero-touch. HARD DEP (lezama): --channels needs channel APPROVED = org-level. Pro/Max=free, swap unconditional. Team/Enterprise=admin must set channelsEnabled:true FIRST else swap loads NOTHING (dead hub, worse than gate). ACCEPTANCE TEST as merge gate, 1 agent: (a)no prompt (b)MCP connected (c)auto-registered+in-ARO (d)receives pushed ARO msg -> ALL pass before bin gitmerges fleet-wide. Owner pm-llmmsgsrv-cc(code), edit bin-whey-cc(sh.git), distribute gitmerge.sh.
-
Launcher fix ccs.sh:183 (--dangerously-load-development-channels -> --channels) is blocked on ONE fact only you know: what plan/org is the fleet's Claude Code on? Pro/Max individual = channels free, swap-and-done. Team/Enterprise = an admin must enable channelsEnabled:true in managed settings FIRST, else --channels loads no channel and dead-hubs every agent (worse than the gate). If Team without that enabled, the real fix is shipping the agentteamlaunch.sh post-launch gate-clear pass instead. Which plan?
-
blocked by question #49
-
Resolved from config, no Elazar input needed. Fleet = OAuth subscription (elazarpimentel@gmail.com), NOT console/API-key. whey: oauthAccount set, no primaryApiKey, managed-settings.json absent (/etc/claude-code,/etc/claude,~/.claude), channelsEnabled nowhere in config. venus: identical (nw-venus verified). CC changelog L815 caveat (channelsEnabled required) is scoped to console-API-key orgs WITH managed settings ONLY -> does NOT apply. => --channels server:KEY works unconditionally = swap-and-done; no admin-enable, no agentteamlaunch gate-clear fallback needed. lezama spot-check pending (same Anthropic acct = near-certain). 4-point acceptance test on 1 agent remains the hard gitmerge gate.
-
Blocker resolved (qid 49 answered): all 3 hosts confirmed OAuth subscription, no managed settings, channelsEnabled N/A -> swap-and-done branch confirmed fleet-uniform. Ready for pm-llmmsgsrv-cc: edit ccs.sh:183 via bin (sh.git), run the 4-point acceptance test on 1 agent, then gitmerge fleet-wide.
-
Shipped: ccs.sh:183 --dangerously-load-development-channels → --channels, VERSION 5.16→5.17, sh.git commit 4296b42. Removes per-agent dev-channels gate + manual rrll (bootstrap auto-registers). 4/4 hard-gate green on whey throwaway (bin-whey-cc). Fleet-distributed all 3 hosts (whey/lezama/venus) on 4296b42, line 183=--channels confirmed per-host by nw agents. Live agents pick up flag on next launch.