The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied academic papers and manuscripts. While this presents an attack surface for indirect prompt injection, the risk is mitigated by explicit 'HARD RULES' that restrict the agent to rhetorical analysis and prevent the reproduction or execution of content from the source files.
[COMMAND_EXECUTION]: The skill employs the mineru tool via a python3 subprocess to perform PDF-to-Markdown conversion. This is a functional requirement and is performed only on user-specified local files.
[EXTERNAL_DOWNLOADS]: The skill mentions external GitHub repositories for optional writing rules but requires the user to provide local paths, preventing unauthorized network downloads.

journal-adapt