telemetry: add cli.run completion event with exit code — measure run success rate #688
Description
Problem
We emit cli.run when a user starts a squad run but have no signal for whether it succeeded or failed. The activation funnel has a blind spot between "ran" and "got value":
install → init → cli.run → ??? → D1 return
Without a completion event, we cannot tell whether users who churned ran successfully or hit errors.
Proposed event
{
"event": "cli.run.complete",
"properties": {
"exit_code": 0,
"duration_ms": 12345,
"agent_count": 1,
"had_error": false
}
}Why it matters
- Distinguish "tried but failed" from "got value" in the activation funnel
- Identify the most common failure modes (errors, timeouts, config issues) to prioritize fixes
- Measure success rate trends over releases — know if a new version improves or degrades activation quality
- The gap between run-started and D1-retained users is currently unmeasurable
User journey stage filled
install → init → cli.run → cli.run.complete (success/fail) → D1 return
Related
- squads-cli telemetry: add cli.install postinstall event to close npm-to-run activation gap #670 (cli.install postinstall event)
- squads-cli telemetry: capture install source / referrer on first run (acquisition blind spot) #655 (install source/referrer)
- squads-cli telemetry: BQ ingestion pipeline not receiving CLI events #687 (BQ pipeline fix — prerequisite for all telemetry analysis)
Source: analytics squad — blind spot identified from activation funnel analysis