The Agent Skills Directory

[SAFE]: The skill instructions and associated files demonstrate safe behavior focused on academic writing support. No indicators of obfuscation, unauthorized data access, or persistence were found. The skill references the author's own GitHub repository (github.com/aipoch/medical-research-skills) for documentation, which is a standard vendor resource.
[COMMAND_EXECUTION]: The provided Python script (scripts/main.py) is a benign CLI utility for generating document outlines based on hardcoded templates. It allows users to optionally output the outline to a file, which is a standard functional feature and does not pose a security risk.
[PROMPT_INJECTION]: The skill processes user-supplied research findings. Analysis of this Indirect Prompt Injection surface shows: (1) Ingestion point: Step 1 (Collect Inputs) gathers research findings and hypotheses; (2) Boundary markers: Present in the form of citation placeholders ([CITE: ...]) and a mandatory Step 3 revision checklist; (3) Capability inventory: The agent generates prose and the included Python script can write to a file; (4) Sanitization: No explicit sanitization of input findings is performed. The risk is assessed as safe due to the highly structured nature of the intended output and the provided workflow constraints.

discussion-composer