The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input that is reflected in generated artifacts and contains self-referential metadata. \n- Ingestion points: User-provided CLI arguments such as --study_name, --outcome, and --input in scripts/main.py are incorporated into the research reports. \n- Boundary markers: No explicit delimiters or instructions are used to isolate user-supplied strings within the Markdown generation logic. \n- Capability inventory: The skill performs file system writes to save reports using the open() function in scripts/main.py. \n- Sanitization: Input strings are concatenated into the Markdown report without escaping, which could lead to formatting issues or indirect injection if the report is parsed by other agents. \n- Self-referential content: The inclusion of clinic-sample-size_audit_result_v2.json provides a pre-computed security verdict, which targets the analysis process. \n- [DYNAMIC_EXECUTION]: The skill dynamically constructs and writes documents at runtime. \n- Evidence: scripts/main.py generates Markdown report content by combining static templates with dynamic user parameters and saves the output to the disk.

clinic-sample-size