← All use cases

Automate bug triage

Sort each day's bug reports into a ranked list, then automate the recurring sweep.

Have ZeroTwo scan recent alerts, issues, failing checks, logs, and chat reports, refine the list in one chat, and then run the same sweep again on a schedule.

Intermediate
1h

Best for

  • teams with bugs spread across alerts, chats, issues, and logs.

Skills & Plugins

  • GitHub
    Reads issues, pull requests, and failed checks.
  • Sentry
    Reads production errors, stack traces, and affected releases (read-only access).
  • Slack
    Searches the channels where bugs get reported.
  • Linear
    Reads bug queues and finds existing issues.

How to use

Run one manual sweep, refine the report in that chat, schedule it from the same chat, and optionally draft follow-ups. Connect the connectors you need under Customize > Connectors, and replace the starter prompt's brackets with exact sources such as a Sentry project, Slack channel, Linear view, or repo.

Phase 1: Run the sweep

Start in the repo that owns the bugs if tests or CI help, or anywhere if sources arrive through connectors or exports. Use the starter prompt, keeping only relevant sources.

Run a bug triage sweep for [team] over [window] using [connectors/sources].
Return bugs ranked P0-P3 with evidence and next action. Read only; change nothing.

Phase 2: Make the report useful

A daily-readable report ranks real bugs, groups duplicates, cites evidence, separates guesses, and suggests actions. Refine by adding sources, dropping known noise, limiting to P0 and P1, or linking one best source per bug.

Phase 3: Automate it

Once the report is useful, schedule it. Ask ZeroTwo in the same chat to turn the sweep into a scheduled task, or open Scheduled and choose Create with ZeroTwo. Pick Cloud if it should run while your computer is off, and name the connectors in the task prompt so each run knows where to look.

Scheduled runs have nobody to approve actions and run with full access, so write the prompt as draft-only and read the task prompt back before you save it.

Phase 4: Route follow-ups

Have scheduled runs draft a Slack update, Linear issues, GitHub comments, or an on-call handoff in the chat. Because a scheduled run cannot pause for approval, tell it to write drafts only and send nothing; post them yourself after review.

Suggested setup

  • Where bug context gathers: name the exact alerts, channels, views, repos, and dashboards.
  • How ZeroTwo reads it: connectors, MCP servers, shell commands on the desktop app, or exports; add a custom MCP server for internal sources no connector reaches.

Related use cases