DOCX/PDF Resume Manipulation Test Suite

I want to push our resume-parsing engine to its limits by feeding it deliberately tricky files. I will send you a clean résumé in plain text; your job is to engineer multiple adversarial versions of it—both DOCX and PDF right from the start—using every document-level sleight of hand you know. Hidden text, layered images covering text, font or glyph substitution, OOXML rewrites that mis-align the visual and logical order, even extraction quirks that only show up when the file is flattened to PDF: anything that could fool an ATS should appear somewhere in the series.

I am flexible on tooling—Python, C#, or a mix—so choose whichever lets you script the generation process cleanly and document it for repeatability. The end goal is twofold: realistic files that open without warning in Word/Reader, and a transparent methodology I can re-run as we harden our software.

Please provide:

• A stepped set of adversarial resumes (easy, moderate, hard, extreme) in both DOCX and PDF
• Well-commented source code and any helper scripts that create or post-process the files
• A concise manifest mapping each file to the exact manipulation techniques used
• Notes on test methodology so my team can extend the suite later

I’ll review portfolios that show prior work with DOCX/PDF programmatic generation, OOXML hacking, PDF object editing or similar parser-focused QA. When you reply, include a short summary of relevant projects, a rough timeline for producing the first batch, and your fixed cost or milestone breakdown. Let’s make our parser bullet-proof.