Category A: Skip present_report reviewer when DocBaker validation_report.json already passed

[sansari]

Parent: #405

Problem

When the Presentation Expert generates a DocBaker HTML report, two layers of validation run back-to-back:

1. DocBaker's own validation — headless Chrome renders the document, takes screenshots of every section and every chart, and writes a validation_report.json with status, errors, warnings, and a full screenshot list.
2. deepwork:reviewer — "Review present_report output" — a separate agent is then spawned that re-reads those same screenshots to evaluate quality.

When validation_report.json shows "status": "pass" and "errors": [], the external reviewer adds no new information. It only re-confirms what the machine already verified.

Evidence

From session logs (sansari-terminal-may12.txt):

Review	Time	Tokens
Review present_report (first DocBaker pass)	4m 6s	50.7k
Review updated present_report (second DocBaker pass)	3m 39s	47.5k

Both ran immediately after DocBaker reported "status": "pass". The GSC report was run through DocBaker twice (after user edits), so this pattern consumed ~8 minutes for a single report.

Proposed fix

Before spawning the present_report reviewer, check validation_report.json:

if validation_report["status"] == "pass" and validation_report["errors"] == []:
    # skip reviewer, mark step as passed directly
    mark_review_as_passed(review_id)
else:
    # run reviewer normally
    spawn_reviewer(...)

If validation_report.json is absent or shows errors/warnings, fall through to the full reviewer as today.

Quality preservation

Full reviewer still runs when:

• DocBaker validation failed or produced warnings
• validation_report.json is missing
• The report format is not DocBaker (Slidev, plain Markdown, etc.)

The reviewer is only skipped when the machine validator already issued a clean pass — the case where it adds the least marginal value.