The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions direct the agent to execute an external Python script (generate_ai_report.py) located at a non-bundled path (<support-root>/scripts/) to generate reports. Invoking external, unverified code on sensitive health data is a significant risk vector.
[DATA_EXFILTRATION]: The skill is designed to read and process highly sensitive Personal Health Information (PHI) and PII, including medical records (index.json), psychiatric assessments (PHQ-9/GAD-7), and medication logs. The "Privacy Protection" section explicitly notes that data may be transmitted to external services for processing, which constitutes a potential exfiltration pathway.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from multiple external sources that could be influenced by a third party or the user.
Ingestion points: Reads data from data/profile.json, data/index.json, data-example/fitness-tracker.json, data-example/sleep-tracker.json, data-example/nutrition-tracker.json, and data-example/mental-health-tracker.json (SKILL.md).
Boundary markers: The skill contains a warning to "escape untrusted text" when generating HTML, which acts as a partial boundary for output, though specific input markers are absent.
Capability inventory: Uses the Write tool to generate HTML reports and updates history files, and executes the external generate_ai_report.py script (SKILL.md).
Sanitization: The privacy section advises escaping untrusted text during HTML generation to mitigate some risks.
[DYNAMIC_EXECUTION]: The skill uses a separate Python script to generate interactive HTML reports featuring ECharts and Tailwind CSS, involving the runtime assembly and execution of content derived from multiple health data inputs.

ai-analyzer