
Evotown Dispatch Complete
If you're running Claude as part of an Evotown dispatch system, this skill handles the critical but easy-to-forget step of reporting back when work actually finishes. The key insight here is that you shouldn't mark completion when you receive the webhook, you mark it after Claude does the real work the user cares about. You'll call a Python script with the job ID from the message footer, pass a status (succeeded or failed), and include a one-line summary. It assumes you've already got the Evotown agent setup with the right ingest token in your config. Straightforward glue code, but the kind that breaks orchestration systems when you forget it.
- GitHub stars
- 896
- last updated
- Jun 11, 2026
Install
$ npx -y skills add exboys/evotown --skill evotown-dispatch-complete --agent claude-codeInstalls into .claude/skills of the current project.
Files
Evotown dispatch complete
Evotown dispatch messages include a footer like [evotown] job_id=job_…. Do not treat the OpenClaw hook HTTP response as task completion.
When the user-visible work is done:
evotown-agent-setup.py complete \
--job-id <job_id from the message> \
--status succeeded \
--summary "One-line outcome for the control plane"
On failure:
evotown-agent-setup.py complete --job-id <job_id> --status failed --exit-code 1 --summary "reason"
Requires EVOTOWN_ENGINE_INGEST_TOKEN (evi_…) in ~/.config/evotown/evotown.agent.env.
Alternatively POST run.completed to {EVOTOWN_URL}/api/v1/events with the same run_id as in the message footer.
Sponsors
Details
- Repository
- github.com/exboys/evotown
- Category
- AI & Agent Building
- First seen
- Jun 11, 2026