The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted health data (CSV, JSON, Apple Health XML) from external sources. These ingestion points could contain malicious instructions designed to influence the agent's behavior during the analysis or recommendation phases.
Ingestion points: readAppleHealthExport(exportPath), readGenericFile(filePath), and glob('data/blood-pressure/**/*.json').
Boundary markers: Not explicitly defined in the provided logic for data parsing.
Capability inventory: The skill uses readFile, writeFile, and glob to manage health data.
Sanitization: Basic data type validation is performed (e.g., parseInt, validateImportedData), but no specific protection against embedded natural language instructions in data fields is present.
[EXTERNAL_DOWNLOADS]: The skill references Python scripts that utilize the requests library to fetch data from Fitbit and Oura Ring APIs. While these are well-known services, they involve downloading external content into the agent's local environment.
[COMMAND_EXECUTION]: The detailed guide describes the execution of Python scripts (e.g., python scripts/import_apple_health.py) for data processing. While these scripts appear to be part of the skill's intended functionality, they represent a path for local code execution based on provided file inputs.

wellally-tech