The Agent Skills Directory

[SAFE]: The skill's code and instructions do not contain any malicious patterns such as credential theft, data exfiltration, or unauthorized command execution. The logic and metadata are consistent with the stated purpose of academic document analysis.
[COMMAND_EXECUTION]: The skill utilizes shell execution capabilities to run PDF parsing scripts.
Evidence: The allowed-tools frontmatter in SKILL.md includes Bash(python:*) and Bash(uv:*). Documentation in references/pdf_explore_usage.md confirms the use of these tools for document processing.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it is designed to ingest and process untrusted PDF data from external sources.
Ingestion points: PDF documents processed as described in SKILL.md.
Boundary markers: No specific delimiters or "ignore instructions" markers are present to prevent the agent from potentially obeying malicious instructions embedded within a parsed PDF.
Capability inventory: The skill is configured to use Read, Write, Edit, and Bash (via python and uv) according to the allowed-tools metadata.
Sanitization: There is no evidence of content filtering or sanitization being applied to the text extracted from PDFs before it is interpreted by the agent.

alterlab-pdf-explore