The Agent Skills Directory

[COMMAND_EXECUTION]: The skill explicitly instructs the agent to assume sudo privileges for system-level tasks. It directs the agent to install packages globally (npm install -g puppeteer) and manage background processes using pkill and nohup.\n- [EXTERNAL_DOWNLOADS]: The skill triggers the automated installation of pymupdf and puppeteer from public registries. Although these are well-known libraries, the instruction to install them immediately with elevated privileges and without version pinning increases the system attack surface.\n- [REMOTE_CODE_EXECUTION]: The requirement to install and execute system-wide packages at runtime facilitates potential remote code execution from external sources.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its data ingestion workflow:\n
Ingestion points: Untrusted data enters the agent context through PDF text extraction via PyMuPDF (described in SKILL.md).\n
Boundary markers: There are no defined delimiters or warnings to ignore instructions embedded within the extracted schematic text.\n
Capability inventory: The agent possesses extensive capabilities including bash command execution, sudo access, and file system write access.\n
Sanitization: The skill lacks logic to sanitize, escape, or validate the content extracted from PDF files before processing.

schematics-pdf-analyser