The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes health data from potentially untrusted sources like symptoms, mood logs, and diet records to generate reports. While central to its function, this presents a surface for indirect prompt injection if malicious content is embedded in source logs.
Ingestion points: Data is read from multiple JSON file paths including data/symptoms/**/*.json, data/mood/**/*.json, and data/diet/**/*.json using Read and Glob tools.
Boundary markers: The instructions do not specify the use of clear delimiters or explicit instructions to ignore embedded commands when interpolating log data.
Capability inventory: The skill has Read access to health records and Write access to create HTML reports in the data/health-reports/ directory.
Sanitization: There is no evidence of data sanitization or escaping logic to prevent malicious payloads in logs from affecting the agent's behavior or the rendered HTML output.
[EXTERNAL_DOWNLOADS]: The skill's HTML report template (templates/report-template.html) references external scripts and styles from well-known and trusted services.
Evidence: The template fetches Tailwind CSS from cdn.tailwindcss.com, Lucide icons from unpkg.com, and the ECharts library from cdn.jsdelivr.net.
Note: These references are used solely for frontend rendering of the generated health reports and target well-known, established services.

health-trend-analyzer