The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill is designed to fetch official documentation live from getstream.io using the .md URL convention (e.g., https://getstream.io/chat/docs/sdk/ios/basics/integration.md). This is a trusted source and the core mechanism of the skill.
[COMMAND_EXECUTION]: The skill generates shell commands for project initialization, environment configuration, and credential management using the getstream CLI. These are standard developer operations and include explicit warnings against hardcoding secrets.
[INDIRECT_PROMPT_INJECTION]: As the skill ingests external data (official documentation and potentially UI reference links), it possesses a theoretical attack surface for indirect prompt injection. However, it implements boundary markers (citations) and enforces strict capability limits (no secret exposure, no arbitrary execution of remote scripts). The risk is internally assessed as low.
[CREDENTIALS_UNSAFE]: The skill explicitly forbids hardcoding API secrets and provides instructions for managing credentials via environment variables or a dedicated Secrets.xcconfig file generated by the CLI, adhering to secure development practices.

stream-swift