AI Complexity agent

Repeat-Failure Learnings

Details

Runs on every pull request, on GitHub

Agent Prompt

Reviews a GitHub PR for recurrences of failure modes the team keeps re-patching. Judges the diff against the project's accumulated review learnings plus a curated baseline of recurrence classes, and flags any diff line that reintroduces one. Use when a PR touches code in areas the team has fixed and re-fixed before.

Ask for it

Review PR {number} for repeat failures

Sample Output

What a Repeat-Failure Learnings report looks like.

IonWarpbotcommented 2 min ago

IonWarp Review — PR #337

3 issues| 2 P1 1 P2

#SevIssueFile

1Bare catch reintroduced — 4th recurrencesync/worker.ts

2Fix shipped as opt-in flag on stored statelib/rollout.ts

3Per-instance patch; sibling sites keep the bugqueue/drain.ts

sync/worker.ts

545556

−++

try { await push(row); } catch (error) { /* dropped */ }try { await push(row); } catch (error) {  throw new PushFailed(row.id, error); // learning: swallowed-sync-error

Reintroduced a learning the team fixed three times · Repeat-Failure Learnings

This exact swallowed-sync-error shape was re-patched in #212, #260 and #291. The cached learning says: rethrow named, never drop.

Suggested fixAI fix prompt

Repeat-Failure Learnings

Details

What a Repeat-Failure Learnings report looks like.

IonWarp Review — PR #337

Reintroduced a learning the team fixed three times · Repeat-Failure Learnings

More agents in the swarm.