Phase 3: Review Curated Outputs

The downstream analysis agents depend entirely on curated content quality — poor inputs produce poor analysis. This phase is the team’s quality gate before automated analysis begins.

Reviewing HTML Mockups

Check that each HTML mockup meets the following criteria:

• Faithful representation — the mockup accurately represents the original screenshot.
• Text accuracy — all visible text has been captured correctly.
• PII replacement — any PII visible in the original has been replaced with fake equivalents.
• UI structure — forms, tables, navigation elements, and layout are correctly represented in the HTML.
• Completeness — no screens have been missed. Cross-reference the html/ directory against the screenshots/ directory to confirm every screenshot has a corresponding mockup.

Reviewing Curated Transcripts

Check that each curated transcript meets the following criteria:

• Relevant content preserved — domain knowledge and application walkthroughs are preserved verbatim.
• Off-topic content removed — project management discussions, scheduling, and social chat have been stripped out.
• PII-free — no personally identifiable information remains in the curated transcript.
• No accidental removal — important context has not been accidentally removed. Compare against the original transcript if anything seems incomplete.

Who Does the Review

The BA/UR on the team is best placed to lead this review, as they have the domain context from conducting the stakeholder interviews. They can most readily spot missing content or inaccurate representations.

Outcome

Confidence that curated outputs are ready for the automated analysis pipeline. If issues are found, re-run the relevant skill on the affected files:

• For a flawed HTML mockup, re-run image-to-html on the original screenshot.
• For a flawed curated transcript, re-run curate-transcript on the original transcript.