The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted PDF files, creating a vulnerability surface for indirect prompt injection. In scripts/agent.py, the script extracts raw content like JavaScript snippets and URLs from potentially malicious PDFs and includes them in an analysis report without sanitization or boundary markers to prevent agent misinterpretation. The ingestion points include the file path provided to the script, while capabilities include subprocess execution and local file writes.
[COMMAND_EXECUTION]: The script automates the use of PDF analysis tools by executing them as sub-processes. Evidence: scripts/agent.py uses subprocess.run() to invoke peepdf, pdfid, and pdf-parser.py to process input files.
[EXTERNAL_DOWNLOADS]: The documentation references external sources for analysis dependencies. Evidence: SKILL.md and references/api-reference.md provide instructions to install the peepdf-3 package via pip and clone the source code from the author's public repository (github.com/jesparza/peepdf.git).
[DATA_EXFILTRATION]: The skill references standard network-based security lookups. Evidence: references/api-reference.md describes using the vtcheck command within the peepdf tool to check file hashes against the VirusTotal database, which involves sending hash metadata to an external service.

analyzing-malicious-pdf-with-peepdf