Workstreams · real client evidence
One request becomes a review-ready job.
An employee asks for an outcome in OpenCode. Toolkit selects the standard workstream, runs every safe check locally, packages the report, and stops before a qualified person must review or record the result.
Available now · v1.0.0
Full client website audit
Run the approved end-to-end website audit and package one standard, evidence-backed client deliverable.
2 real client runs · 58 locally recorded events
2 clients. The same company standard.
These are sanitized recordings of OpenCode sessions against client domains explicitly selected for this internal proof—not a scripted fixture and not a forced green demo.
Rau Plastic Surgery
rauplasticsurgery.com

Real read-only OpenCode run against rauplasticsurgery.com. No client configuration or content was changed; the audit made normal web requests that may appear in analytics. The run stopped before peer review, approval, ticket updates, or Activity Logs write-back.
Captured Sep 5, 2026 · OpenCode · work item DEMO-RAU-RECAPTURE-004
Integrity check: 18/18 file references matched their recorded SHA-256 digests.
Sherber & Rad
sherberandrad.com

Real read-only OpenCode run against sherberandrad.com. No client configuration or content was changed; the audit made normal web requests that may appear in analytics. The run stopped before peer review, approval, ticket updates, or Activity Logs write-back.
Captured Sep 5, 2026 · OpenCode · work item DEMO-SHERBER-002
Integrity check: 20/20 file references matched their recorded SHA-256 digests.
What the runs actually found
A failed check is useful evidence. Toolkit preserves it, packages it, and routes it for review.
Rau Plastic Surgery
162 URLs discovered · 6 quality lanes · 3 browser-QA pages.
- Local browser: desktop PASS · iphone-17 PASS
- PageSpeed: mobile 36 · desktop 56
- Analytics: REVIEW · published GTM: NOT_APPLICABLE
- 6 normalized records · JSON, HTML, PDF
Sherber & Rad
707 URLs discovered · 8 quality lanes · 3 browser-QA pages.
- Local browser: desktop PASS · iphone-17 FAIL
- PageSpeed: mobile 44 · desktop 60
- Analytics: REVIEW · published GTM: REVIEW
- 7 normalized records · JSON, HTML, PDF
The standard lifecycle
Both sessions followed this same version-pinned definition. The first nine steps resolved through automation or an explicit non-applicable reason; the final two remained human-controlled.
- 01Confirm scope, client, and work recorddone3 evidence
- 02Discover the website and select representative pagesdone2 evidence
- 03Run the full site quality and technical checklistdone2 evidence
- 04Measure local browser performancedone2 evidence
- 05Measure PageSpeed and real-user performancedone2 evidence
- 06Inspect Cloudflare and edge configurationdone2 evidence
- 07Audit analytics and tag coveragedone2 evidence
- 08Audit published Google Tag Manager behaviornot applicable — the analytics audit found no Google Tag Manager container, no GTM tag, and no dataLayer across 162 discovered pagesskippedreason recorded
- 09Check the AI-crawler discovery filenot captured — this required coverage step was added after the recorded read-only sessionskippedreason recorded
- 10Generate the standard client audit reportdone3 evidence
- 11Qualified peer reviews coverage, findings, and deliverablesreadyhuman gate
- 12Write completion and evidence links to the system of recordpendingafter review
Where automation stops
automated locally
- Discovery, QA, PageSpeed, analytics, GTM, and edge checks
- Evidence hashing and standard reports
- Resume after interruption
qualified peer
- Coverage and findings review
- Deliverable approval
- Ticket or Activity Logs write-back
human locked
- DNS and cutovers
- Publishing and external sends
- Destructive changes
Start with the safe plan
The friendly Toolkit command performs no checks and makes no changes. Starting a run also requires a named employee and an originating ticket or Activity Logs fallback.
./toolkit run "Full client website audit" --url https://sherberandrad.com/Windows uses .\toolkit.ps1 with the same friendly name and arguments.