test(polling): add Phase 4 polling workflow integration tests #483

Merged

clawdie merged 1 commit from test/phase4-polling-workflow into main

2026-08-24 06:00:12 +02:00

clawdie commented

2026-08-24 05:56:18 +02:00

Owner

Three tests simulating the remote agent cron loop (colibri_poll.py + colibri_task_done.py):

polling_workflow_capability_routing — two agents with disjoint capabilities,
intake tasks routed by scheduler, each agent polls for its tasks and
transitions to done. Proves end-to-end agent-side capability routing.
polling_workflow_single_agent_multiple_tasks — one agent with multiple
capabilities claims both tasks sequentially. Documents current contention
behavior (no guard) and session isolation across polling cycles.
polling_workflow_no_matching_capability — agent with non-matching capability
sees empty list. Verifies filtering doesn't leak tasks.

All tests run on Linux (pure Unix socket, no FreeBSD deps).

Three tests simulating the remote agent cron loop (colibri_poll.py + colibri_task_done.py): 1. polling_workflow_capability_routing — two agents with disjoint capabilities, intake tasks routed by scheduler, each agent polls for its tasks and transitions to done. Proves end-to-end agent-side capability routing. 2. polling_workflow_single_agent_multiple_tasks — one agent with multiple capabilities claims both tasks sequentially. Documents current contention behavior (no guard) and session isolation across polling cycles. 3. polling_workflow_no_matching_capability — agent with non-matching capability sees empty list. Verifies filtering doesn't leak tasks. All tests run on Linux (pure Unix socket, no FreeBSD deps).

clawdie added 1 commit

2026-08-24 05:56:24 +02:00

test(polling): add Phase 4 polling workflow integration tests

CI / rust (pull_request) Failing after 39s

Details

CI / python-tests (pull_request) Successful in 49s

Details

CI / markdown (pull_request) Failing after 33s

Details

format-check / prettier (pull_request) Failing after 45s

Details

CI / port (pull_request) Failing after 28s

Details

leak-guard / guard (pull_request) Successful in 1m20s

Details

CI / agent-jail-pkgs (pull_request) Failing after 23s

Details

ci-gate / gate (pull_request) Successful in 4m49s

Details

c56e83665e

Three tests simulating the remote agent cron loop (colibri_poll.py + colibri_task_done.py):

1. polling_workflow_capability_routing — two agents with disjoint capabilities,
   intake tasks routed by scheduler, each agent polls for its tasks and
   transitions to done. Proves end-to-end agent-side capability routing.

2. polling_workflow_single_agent_multiple_tasks — one agent with multiple
   capabilities claims both tasks sequentially. Documents current contention
   behavior (no guard) and session isolation across polling cycles.

3. polling_workflow_no_matching_capability — agent with non-matching capability
   sees empty list. Verifies filtering doesn't leak tasks.

All tests run on Linux (pure Unix socket, no FreeBSD deps).

clawdie merged commit 7017451338 into main

2026-08-24 06:00:11 +02:00

clawdie referenced this pull request from a commit

2026-08-24 06:00:14 +02:00

Merge pull request 'test(polling): add Phase 4 polling workflow integration tests' (#483) from test/phase4-polling-workflow into main

clawdie deleted branch test/phase4-polling-workflow

2026-08-24 06:01:13 +02:00