Real client audit: Sherber & Rad

Watch OpenCode run the approved audit against Sherber & Rad, package the evidence, and stop at qualified peer review.

The OpenCode clip loops continuously. The complete text transcript is below.

What this session proves

OpenCodeClient auditEvidencePeer review

Ask for a clear outcomeName the target, checks, evidence and output. The toolkit can then select the right repeatable process.

Review before anything shipsUse the returned evidence to verify the work. A report or email draft is a handoff for review, not permission to send.

Improve repeat workCapture missing fields, screenshots and approval steps. A repeated job is a candidate for a stronger shared skill.

Try the first prompt

This is the exact employee-facing request used for this lesson.

Audit sherberandrad.com with the Full client website audit workstream. Read-only; package the report and stop before peer review or write-back.

Results transcript

The complete text result is always available with the recording.

OpenCode · build · deepseek-v4-flash

Request accepted
  Workstream: Full client website audit v1.0.0
  Target: https://sherberandrad.com
  Mode: read-only
  Work item: DEMO-SHERBER-002

Running the approved lifecycle…
  ✓ Confirm scope, client, and work record
  ✓ Discover the website and select representative pages
  ✓ Run the full site quality and technical checklist
  ✓ Measure local browser performance
  ✓ Measure PageSpeed and real-user performance
  ✓ Inspect Cloudflare and edge configuration
  ✓ Audit analytics and tag coverage
  ✓ Audit published Google Tag Manager behavior
  ✓ Generate the standard client audit report

OpenCode summary · sherberandrad.com
  9 automated steps completed
  30 append-only lifecycle events
  20 hashed evidence references
  157 local artifacts · JSON, HTML, PDF

Actual results
  Full site quality: FAIL · 707 URLs discovered · 8 lanes
  Local browser: desktop PASS · iPhone LCP FAIL
  PageSpeed: mobile 44 FAIL · desktop 60 REVIEW
  Analytics coverage: REVIEW
  Published GTM behavior: REVIEW
  Standard report: FAIL · 7 normalized records

Qualified peer review: READY
Write-back: PENDING behind review

Stopped here as required.
No approval requested. No client system changed. Nothing published or written back.

Keep it safe and useful

Start a new task when the work changes, protect keys and .env files, and stop a run that is heading in the wrong direction. A person remains the final reviewer before a client-facing or live-system change. See API-key hygiene.

If this reveals a useful gap, submit an issue. The next recording and the next run should benefit from what this one taught.