Bug Vigil

How this loop works

Suggested schedule

Patrol cadence (you set it)

Loop type

Closed loop — goal-bound; it finishes itself when the goal is met.

Exit condition

Finishes on one clean capture of the bug (or a bounded give-up window).

Notifications

On by default — you're alerted on each run's outcome (completion or failure). Silence with notify: never.

Ease of starting

Some setup and context are needed before the first useful run.

When you see value

Waits out an intermittent bug; the capture can take days or weeks until it recurs.

Set up a closed loop that stakes out one specific hard-to-reproduce bug until it recurs with enough evidence to fix it. Describe the bug with me first - the symptom, where it shows up, what makes it intermittent - and before creating anything, verify a concrete way to observe a recurrence exists on this machine - an error tracker, logs, a metric or query, a reproducible check, or gh - and run it once as a smoke test; if a recurrence cannot actually be detected, say so and do not create a blind loop. Define the finish condition with me - typically one clean capture of the bug with enough context to fix it, or a bounded give-up window - and create the loop closed, with that as its goal. On the cadence we agree, patrol the observation path for a recurrence. When it fires, capture the full context in one dated report with front matter `type: report`, a `title`, and `date` - the stack trace, the session or request id, the surrounding events and timing, the inputs and state - everything a fix would need, and never copy credentials or personal data into it. Then finish the loop, its goal met. If the window closes with no recurrence, report that plainly and finish. Set up a modest dashboard at creation: an embed of the latest capture (or the current patrol status) so the evidence is one click away.