pdf-ocr
Fail
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: HIGHPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill instructions direct the agent to use
sudo apt-get installto install Tesseract OCR, which involves acquiring root-level permissions. - [COMMAND_EXECUTION]: The environment setup requires the execution of multiple shell commands using
apt-get,brew, andpipto install necessary system and Python dependencies. - [INDIRECT_PROMPT_INJECTION]: The skill ingests text from external, untrusted PDF documents, creating a vulnerability where hidden instructions in a document could manipulate agent behavior.
- Ingestion points: User-provided PDF files are parsed and OCR'ed as described in Step 1 and Step 3 of
SKILL.md. - Boundary markers: Absent. The instructions do not include markers or directives to treat extracted text as untrusted or to ignore embedded instructions.
- Capability inventory: The skill environment allows for shell command execution (package managers) and file system writes (
save_resultsfunction in Step 6). - Sanitization: Absent. The extracted text is saved directly to output files without any validation or sanitization logic to detect or strip malicious content.
Recommendations
- AI detected serious security threats
Audit Metadata