doc-redact-page-review

Orchestrate two-pass page-level OCR/CSV redaction review and merge.

55|11|Updated Apr 25, 2024
One-click install
npx skills add https://github.com/seanpedrick-case/doc_redaction --skill doc-redact-page-review
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: doc-redact-page-review
Source: https://github.com/seanpedrick-case/doc_redaction/tree/main/skills/doc-redact-page-review
Command: npx skills add https://github.com/seanpedrick-case/doc_redaction --skill doc-redact-page-review

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps you review and correct OCR-based redactions page-by-page, so you can confidently apply accurate removals/additions across a selected page range without corrupting the rest of the document.

Core Features & Use Cases

  • Parallel Pass 1 page review (OCR/CSV only): Spawns one child per page to inspect OCR line/word outputs, apply review rules, and return page-scoped verified row deltas.
  • Master CSV-safe merge: Loads a single master review CSV once, preserves untouched pages, and replaces only the reviewed range with child-verified rows, then sorts and validates the merged output.
  • Single, controlled apply cycle: Runs coverage verification, optional suspicious-row pruning, previews, and performs exactly one /review_apply for the entire merged/pruned CSV.
  • Optional Pass 2 visual QA with VLM: When requested (or when policy flags demand it), checks only flagged pages visually using sequential VLM calls and conservatively patches CSV results, followed by one more re-apply if changes occur.

Quick Start

Ask it to “Review pages 4–7 in the redaction output folder, using these rules, run Pass 1 OCR/CSV review only, merge into output_review_final, and apply once.”

Frequently Asked Questions about doc-redact-page-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review OCR redaction outputs for specific PDF pages without corrupting the rest of the document?

You can review OCR redaction outputs by running a parallel page-level review across a selected page range. This process applies review rules per page, safely merges the verified rows into the master CSV, and replaces only the targeted range while preserving untouched pages.

What is the best way to merge per-page CSV redaction deltas into a single master review file?

The best way to merge per-page CSV redaction deltas is to load the master review CSV once, replace the reviewed page range with child-verified rows, and sort and validate the merged output. This ensures untouched pages remain intact during the document redaction process.

Can I perform a visual QA check on flagged PDF redaction pages using VLM?

Yes, you can perform a visual QA check using VLM. After the initial OCR/CSV review and apply cycle, an optional Pass 2 checks only flagged pages visually using sequential VLM calls and conservatively patches the CSV results.

How does a two-pass page review workflow improve PII removal accuracy?

A two-pass page review improves PII removal accuracy by first inspecting OCR line/word outputs and returning page-scoped row deltas without calling VLM. The second pass visually verifies only flagged pages, ensuring conservative patching and a single controlled apply cycle.

Do I need a completed redaction run folder to start correcting OCR document redactions?

Yes, you need a completed redaction run folder containing the initial outputs. The workflow requires this existing folder to spawn parallel children for per-page OCR/CSV review, extract deltas, and merge them into a consolidated review file.