AI Bug Fixing & Automated Triage

AI bug fixing with Stoneforge lets you resolve bugs in parallel instead of one at a time. Import your bug backlog, set priorities, and let automated bug triage dispatch multiple agents to investigate and fix bugs simultaneously.

From backlog to fixed with parallel AI agents

Instead of having one developer cycle through bugs sequentially, Stoneforge dispatches multiple bugs to separate agents. Each agent gets its own worktree, reproduces the issue, writes a fix, adds regression tests, and submits a merge request.

# Pull bugs from GitHub Issues
sf external-sync pull --provider github

# Or create bug tasks manually
sf task create --title "Fix: Login fails with SSO redirect loop" --priority 1
sf task create --title "Fix: CSV export truncates long fields" --priority 2
sf task create --title "Fix: Dashboard chart doesn't render on Safari" --priority 2
sf task create --title "Fix: Email notification sent twice on retry" --priority 3

# Agents start fixing bugs in priority order
sf serve

Intelligent automated bug triage

The Director agent can analyze incoming bugs and estimate complexity, identify related bugs that should be fixed together, and flag bugs that might share a root cause. Related bugs are grouped into a single task to avoid duplicate fixes.

Regression tests with every fix

Every AI bug fixing result comes with a regression test. The agent writes a test that fails without the fix and passes with it. The Steward agent runs tests before merging to make sure nothing is broken — similar to the quality checks in automated code review.

Bidirectional sync with issue trackers

When an agent fixes a bug, the linked GitHub Issue or Linear ticket is updated automatically with a summary of the fix, a link to the merge request, and which tests were added.

# Bug fixed → GitHub Issue updated automatically
# Issue comment: "Fixed in MR #142. Root cause: race condition in
# session refresh. Added regression test in auth.test.ts."

Clear the backlog with AI bug fixing

Because agents work in parallel, you can make progress on multiple bugs at once instead of working through them one by one. The medium-priority bugs that tend to sit untouched at the bottom of the backlog finally get attention. Pair this with AI test generation to prevent future regressions.