verify PM loop end-to-end #32

New issue

Closed

opened 2026-08-21 21:30:02 +00:00 by Mark · 1 comment

Mark commented

2026-08-21 21:30:02 +00:00

Owner

PM management-plane end-to-end verification (round 2), performed 2026-08-21.

Card: kanban t_256a84d0 <- Vikunja org_mode #110

Result: PASS — all 8 loop legs observed live through the real gateway dispatcher:

Vikunja task.created -> webhook id 14 -> daemon (org-subscriber.service, active)
branch task.created -> vikunja-webhook-to-kanban.sh (run-dir log 20260821-211947 -> Created t_55c737dd)
handler SSH to blackhole, card created (idempotency key vikunja:113)
card ready on tasker board within ~60s
completion (done=true) -> handler exits "task already done; no card" (by design; outbound reflection owns done state)
reassign card -> manager profile
dispatcher claimed + spawned manager (run 510, pid 1441462)
manager acknowledged + completed card

Defect found (separate from loop): the vikunja-kanban-sync Hermes cron (job b186eb82ff22, every 1m, enabled) still has an inbound create path with its own mapping DB and creates a SECOND card for every actionable Vikunja task that the webhook already carded (#109-#113 all duplicated). Docs (vikunja-pm-system.md §2/§5) say the 1m poll is retired/catch-up only — stale docs + stale cron pairing.

Reference: ~/.hermes/skills/devops/vikunja/references/pm-loop-e2e-2026-08-21.md

Fix direction (NOT executed — verifier only): disable cron job b186eb82ff22, or make vikunja-kanban-sync.py Case 2 dedupe against webhook cards by body match on /tasks/ before creating.

PM management-plane end-to-end verification (round 2), performed 2026-08-21. Card: kanban t_256a84d0 <- Vikunja org_mode #110 Result: PASS — all 8 loop legs observed live through the real gateway dispatcher: 1. Vikunja task.created -> webhook id 14 -> daemon (org-subscriber.service, active) 2. branch task.created -> vikunja-webhook-to-kanban.sh (run-dir log 20260821-211947 -> Created t_55c737dd) 3. handler SSH to blackhole, card created (idempotency key vikunja:113) 4. card ready on tasker board within ~60s 5. completion (done=true) -> handler exits "task already done; no card" (by design; outbound reflection owns done state) 6. reassign card -> manager profile 7. dispatcher claimed + spawned manager (run 510, pid 1441462) 8. manager acknowledged + completed card Defect found (separate from loop): the vikunja-kanban-sync Hermes cron (job b186eb82ff22, every 1m, enabled) still has an inbound create path with its own mapping DB and creates a SECOND card for every actionable Vikunja task that the webhook already carded (#109-#113 all duplicated). Docs (vikunja-pm-system.md §2/§5) say the 1m poll is retired/catch-up only — stale docs + stale cron pairing. Reference: ~/.hermes/skills/devops/vikunja/references/pm-loop-e2e-2026-08-21.md Fix direction (NOT executed — verifier only): disable cron job b186eb82ff22, or make vikunja-kanban-sync.py Case 2 dedupe against webhook cards by body match on /tasks/<tid> before creating.

Mark commented

2026-09-15 15:36:52 +00:00

Author

Owner

Closing — verification card for PM loop, all 8 legs passed 2026-08-21. Paper trail.