The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides and instructs the use of a Python script (scripts/run.py) to handle the execution logic and API interaction.
[EXTERNAL_DOWNLOADS]: The runner script optionally utilizes the certifi Python package to provide a reliable root CA bundle for secure SSL/TLS connections to the API.
[DATA_EXFILTRATION]: User-submitted content from parameters such as materialText, materialUrl, and materialFile is transmitted to the external API endpoint at https://ai-skills.ai for processing.
[CREDENTIALS_UNSAFE]: Authentication with the backend service requires an API key (AISKILLS_API_KEY) which the user must provide as an environment variable.
[PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection because it processes untrusted data from external sources.
Ingestion points: Data enters the system through the materialUrl, materialFile, and materialText input fields.
Boundary markers: The provided script does not implement specific delimiters or safety instructions to prevent the underlying model from following instructions contained within the user data.
Capability inventory: The script's runtime environment is restricted to performing network requests to the vendor's API and displaying the results; it lacks file-writing or arbitrary command execution capabilities.
Sanitization: External inputs are transmitted in a structured JSON format but do not undergo content-level sanitization before being sent to the AI service.

ai-podcast