The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use standard shell utilities such as curl, mktemp, install, and mkdir for legitimate API interaction and local configuration management as part of its onboarding and management functions.\n- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill manages sensitive API tokens with high security awareness. It enforces restricted file permissions (0600) for credentials and mandates the use of the --max-redirs 0 flag with curl to prevent credential leakage through accidental HTTP redirects.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes data from SubRouter API endpoints. Ingestion points occur in the SKILL.md scripts; boundary markers are enforced by instructions to parse specific JSON envelope fields (.data); capabilities are limited to standard configuration tasks; and sanitization is implemented via instructions to filter sensitive secrets from the output displayed to the user.\n- [PERSISTENCE]: The skill stores API credentials in ~/.subrouter/credentials.json to maintain the agent's authorization state across sessions, which is a standard and transparent persistence mechanism for this type of tool.

subrouter