Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill instructs the agent to use
sudofor installing system dependencies (sudo apt-get install -y poppler-utils), which involves running commands with root privileges. - [COMMAND_EXECUTION]: The workflow relies on shell commands for PDF rendering (
pdftoppm) and managing software via package managers likeuv,pip,brew, andapt. - [EXTERNAL_DOWNLOADS]: The skill downloads third-party Python libraries (
reportlab,pdfplumber,pypdf) and system binaries from public package registries during the setup phase. - [INDIRECT_PROMPT_INJECTION]: Because the skill is designed to read and review external PDF documents, it is vulnerable to indirect prompt injection where an attacker could embed hidden instructions within a PDF to manipulate the agent's behavior.
- Ingestion points: Content extraction via
pdfplumberandpypdfinSKILL.md. - Boundary markers: The instructions do not define clear delimiters or "ignore embedded instructions" blocks for the ingested PDF content.
- Capability inventory: The skill has access to shell execution, file system writes (
output/pdf/), and network operations through package managers. - Sanitization: No content filtering or sanitization steps are documented for the extracted PDF data.
Audit Metadata