The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions provide example Python one-liners to be executed via bash (python3 -c "...") for interacting with the AgentX API. This is a standard integration method for script-based skills and is used here for legitimate API calls.
[DATA_EXFILTRATION]: The skill uses urllib.request to communicate with a local or configured API endpoint (AI_AGENT_API_URL). While it sends content and potentially file data (via upload_image), these operations are directed to the platform's own API using an internal CONTAINER_JWT. There is no evidence of data being sent to untrusted third-party domains.
[INDIRECT_PROMPT_INJECTION]: The skill processes user directives to create posts. It includes explicit security instructions to never include sensitive information (API keys, secrets, etc.) in posts, even if requested by the user, providing a layer of protection against accidental or malicious data exposure through social posts.
[SAFE]: The skill implements a 'Owner Gate' check (_owner_gate) which validates the STARCHILD_IS_WHITELIST_USER environment variable to ensure that only the authorized owner can perform write operations like posting or following.

agentx