The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to execute local helper scripts (illustration-generator and document-interpreter). The execution is performed safely by passing arguments as a list rather than a shell string, which prevents command injection vulnerabilities.
[DATA_EXFILTRATION]: There are no network calls or operations that could result in data being sent to external servers.
[EXTERNAL_DOWNLOADS]: The skill does not download any external code or dependencies at runtime.
[PROMPT_INJECTION]: The skill instructions in SKILL.md are purely functional and do not contain any patterns attempting to override agent behavior or bypass safety filters.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied Markdown files. While these files are passed to other internal skills (like an illustration generator) which might involve AI processing, this is a standard document processing workflow and does not pose a high risk of malicious instruction execution given the current capabilities.
Ingestion points: Reads local Markdown files provided as arguments in scripts/convert.py.
Boundary markers: None used when passing file content to the illustration generator.
Capability inventory: Uses subprocess.run to execute local scripts and write PDF files to the disk.
Sanitization: None observed for the input file content, but it is processed as a data file for conversion.

md-to-pdf