The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from an external API, creating a theoretical ingestion surface for indirect prompt injection.
Ingestion points: The scripts/call.py script (line 217) fetches and reads JSON response payloads from the german-tax-calculator.p.rapidapi.com API.
Boundary markers: The skill instructions in SKILL.md strictly limit the agent to reporting the euro amounts and notes returned by the API, providing a logical constraint on how the data is interpreted.
Capability inventory: The skill is authorized to execute its local scripts/call.py script and read provided reference documentation.
Sanitization: The scripts/call.py script validates that the API response is valid JSON before processing it and includes a mask() function to ensure the RAPIDAPI_KEY is never leaked in the output provided to the agent.
[SAFE]: The skill follows security best practices for handling sensitive credentials.
The script scripts/call.py retrieves the RAPIDAPI_KEY from the environment or a .env file, which is the recommended method for secret management.
The script implements a path normalization check (normalize_path) that ensures API calls are restricted to the v1/ endpoint, preventing potential directory traversal or unauthorized endpoint access.

steuerrechner-api