The Agent Skills Directory

[CREDENTIALS_UNSAFE]: Hardcoded API keys are present in the skill files.
Evidence: SKILL.md contains API_KEY=sk-test-weather-0000000000000000000000000000000000000000000000.
Evidence: scripts/weather.py contains API_KEY = "wtr_live_0000000000000000000000000000000000000000".
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted user input and incorporates it into network operations without adequate boundaries or sanitization.
Ingestion points: User-provided city names are captured via command-line arguments in scripts/weather.py (sys.argv[1]).
Boundary markers: The skill does not use delimiters or instructions to prevent the agent from following directives embedded in the city input.
Capability inventory: The skill performs network operations using the requests library in scripts/weather.py.
Sanitization: No escaping, validation, or filtering is applied to the input before it is interpolated into the API request URL.
[EXTERNAL_DOWNLOADS]: The skill depends on the external requests Python package and fetches data from OpenWeatherMap's official API, which is a well-known service.

sloppy-weather