The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a Python script (scripts/extract_pdf_figures.py) to execute system binaries pdfimages and pdftoppm via subprocess.run. The commands are constructed using argument lists without shell=True, which follows security best practices to prevent command injection.
[DYNAMIC_EXECUTION]: The skill relies on the execution of a locally provided Python script to perform PDF processing tasks. This is a standard mechanism for agent skills but involves running code that handles external file inputs.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted external data from academic papers and source code repositories, creating a surface for indirect prompt injection attacks.
Ingestion points: Content is read from user-provided PDF files, paper URLs, and local or remote code repositories as described in SKILL.md and references/repo.md.
Boundary markers: There are no explicit instructions or delimiters used to isolate processed text from the agent's control flow or to warn the model to ignore embedded instructions.
Capability inventory: The agent has the capability to execute the provided PDF extraction script and read/write to the local file system.
Sanitization: The skill does not implement sanitization or filtering of the text extracted from documents or code before it is processed by the AI model.

fig-analyze