The Agent Skills Directory

[SAFE]: The skill facilitates the management of residential proxy configurations and credentials. It provides a structured API in exports.py and CLI scripts for setup, binding, and verification.
[SAFE]: Proxy credentials (usernames and passwords) are persisted in the workspace .env file, following the platform's recommended practices for secret management.
[SAFE]: Network activity is restricted to verifying proxy connectivity via the well-known ifconfig.co service. No sensitive data or telemetry is exfiltrated to unauthorized domains.
[INDIRECT_PROMPT_INJECTION]: The skill provides an API that generates remediation instructions containing shell commands built from untrusted inputs.
Ingestion points: The skill_name parameter in the get_proxy_for_skill and onboarding_guide functions within exports.py.
Boundary markers: No delimiters or escape sequences are applied to the skill name before it is included in the suggested command string.
Capability inventory: The skill possesses capabilities for filesystem access (reading/writing .env and .json files) and network operations (proxy-aware requests).
Sanitization: The skill_name is interpolated directly into a suggested shell command string without validation or shell-escaping. If an automated agent attempts to execute these suggested remediation steps verbatim, a maliciously crafted skill name could result in command injection.

byo-proxy