The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill documentation includes a hardcoded database connection string (mysql+pymysql://tools:tools-password@127.0.0.1:4306/tools_test) within an example command. Although referencing localhost, hardcoded credentials in connection strings are flagged as high-risk findings.
[COMMAND_EXECUTION]: The skill provides specific instructions and examples for the agent to execute shell commands using the pytest tool, which is an intended but powerful capability that should be used with caution.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it ingests user-defined logic to generate and run tests. Evidence Chain: 1. Ingestion points: User descriptions of components to be tested (controllers, models, etc.). 2. Boundary markers: Not present in the provided instructions for generated code. 3. Capability inventory: Shell execution of pytest in the .venv environment. 4. Sanitization: Uses a confirmation step for framework choices but does not validate the logic of the code being tested.

flask-tests