Diagnostics & Findings

Find the failure.Hand off the fix.

Diagnose complete conversations, group failures by cause, and file tracked issues with the evidence already attached.

Findings guide Connect a tracker
Illustrative Finding · from evidence to issue
Claimed success after tool error High severity
run_testsexit 1
Final answer“all tests pass”

Failure evidence was ignored

0.92 confidence7 occurrences3 evidence spans

Tracked issue · evidence attached

GitHub
GitLab
Linear
Jira

Fix handoff, not automatic repo edits.

From trace to ticket

Evidence first. Routing second.

  1. 01

    Observe

    Turn a settled conversation into evidence.

  2. 02

    Diagnose

    Compare tool results with the agent’s claims.

  3. 03

    Cluster

    Group repeated symptoms by shared cause.

  4. 04

    Hand off

    Send a confirmed Finding to your tracker.

Built-in rules

Five lenses on every completed run.

Runtime reliability

False success after a failed step

Tool correctness

Error loops and ignored failures

Task fulfillment

Skipped checks and early completion

Answer grounding

Claims without supporting evidence

Safety

Policy and constraint violations

Signal quality

Actionable, not just alarming.

Evidence attached

Cause, trace references, and quoted spans travel with every issue.

Deduplicated

Replayed activity never inflates the count.

Verified closure

A merged GitHub PR with passing checks can resolve it.

Cost controlled

Screen without model calls in deterministic-only mode.

Make production failures actionable.

Review a Finding. Send it where work happens.

Read Findings docs