1
Brief safely
Start with sanitized context and a concrete reliability decision.
Technical audit
A compact audit before the team commits deeper engineering time.
The audit is useful when a team needs a clear view of architecture risk, missing evaluation coverage, and the next work package before committing to a full reliability sprint or implementation path.
Sanitized brief first
No private repo in form
Artifacts delivered
Ship / revise / stop memo
2
Reproduce failures
Turn traces, examples, or eval runs into repeatable evidence.
3
Define gates
Separate blockers, warnings, thresholds, and ownership.
4
Decide ship / revise / stop
Package the evidence into an engineering-readable memo.
Good fit
- The team needs an independent read before investing in a direction.
- There is enough architecture, repository, paper, trace, or dataset context to review.
- The desired output is a risk memo and prioritized next steps.
Boundaries
- Does not replace a full implementation sprint.
- Does not make compliance or security-control claims without reviewed evidence.
- Does not require sensitive production data in the public intake flow.
Deliverables
- Risk memo
- Evaluation-gap map
- Implementation roadmap
- Recommended next scope
Inspect the artifact shape
The sample is representative structure, not client data. It shows how evaluation plans, taxonomies, gates, and decision memos can be packaged without inventing proof.
View artifact sampleProcess
How the work moves from brief to evidence.
01
Review sanitized context first and define the audit questions.
02
Inspect architecture, workflow boundaries, available evals, and operational risks.
03
Separate immediate blockers from useful improvements.
04
Recommend whether to continue, revise, or defer deeper work.
Next step
Have a workflow ready to evaluate?
Send sanitized context first, then use the call to confirm fit, access boundaries, and scope.
Related technical notes