The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to install well-known Python packages from the official PyPI registry, including Flask, SQLAlchemy, Marshmallow, and Gunicorn. These are standard dependencies for the framework's ecosystem.
[COMMAND_EXECUTION]: The skill describes the execution of standard development and production commands such as flask run and gunicorn. These are the intended methods for operating a Flask application.
[DATA_EXFILTRATION]: The skill provides examples of using the requests library to facilitate communication with external APIs. This is a common pattern for microservices and is used here for legitimate integration purposes.
[PROMPT_INJECTION]: The skill demonstrates patterns for handling untrusted data from web requests which presents a surface for indirect prompt injection. Mandatory Evidence Chain:
Ingestion points: Untrusted data enters the application via request.get_json() and request.args.get() in SKILL.md.
Boundary markers: No specific boundary delimiters are used for the data within the examples, as is typical for web framework documentation.
Capability inventory: The application patterns include file system logging and network operations using the requests library in SKILL.md.
Sanitization: The skill provides comprehensive sanitization and validation patterns using Marshmallow schemas and Pydantic models to mitigate risks.

flask