The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses sensitive personal health information (PHI) stored in local files, including profile.json (containing weight and BMI), hypertension-tracker.json (blood pressure records), and diabetes-tracker.json (blood glucose levels). Although no network exfiltration tools are currently enabled, the access to these specific paths constitutes high-severity data exposure under the analyzer's threat model.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from multiple external sources.
Ingestion points: The agent reads diet and health logs from data-example/nutrition-logs/YYYY-MM/YYYY-MM-DD.json, as well as tracker files for fitness, sleep, and medical conditions.
Boundary markers: The skill definition does not provide explicit delimiters or "ignore embedded instructions" warnings for the agent when processing these files.
Capability inventory: The agent is equipped with Read, Grep, Glob, and Write tools, granting it the ability to read sensitive files and modify the local file system.
Sanitization: The instructions lack any requirement for sanitization, validation, or filtering of the ingested JSON content before it is interpolated into analytical reports.

nutrition-analyzer