ocr-validatelisted

# media-forge:ocr-validate INTERNAL skill. Validates rendered text accuracy in a generated image using Google Cloud Vision OCR. Called exclusively by `media-forge:quality-reviewer`. ## Workflow 1. Receive image path and `expected_text` from caller context. 2. Call Cloud Vision API `annotateImage` with `TEXT_DETECTION` feature on the image. 3. Extract all detected text blocks from the response. 4. Compare detected text against `expected_text` using normalized string comparison (trim, lowercase, collapse whitespace). 5. Return structured result: `{ "pass": boolean, "detected_text": string, "expected_text": string, "mismatch_details": [] }`. ## When to use Called internally by `quality-reviewer` in Stage 1 of the 3-stage review pipeline. Not for direct user invocation. ## Outputs - `pass`: boolean — true if all expected text was detected accurately - `detected_text`: full text extracted by OCR - `mismatch_details`: list of `{ expected, detected, position }` for each discrepancy - Error class on failure: `OCR_MISMATCH`