The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download the Render CLI from the official render-oss GitHub repository (https://raw.githubusercontent.com/render-oss/cli/main/bin/install.sh). This download originates from the infrastructure of a well-known service provider.\n- [REMOTE_CODE_EXECUTION]: The skill documents an installation method that pipes a remote script directly to a shell (curl -fsSL ... | sh). This pattern executes code from a remote source, which in this case is the official repository of a well-known cloud platform.\n- [PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by processing external, potentially untrusted data through task arguments and return values.\n
Ingestion points: Untrusted data enters the agent's context through task input arguments in client calls like run_task and start_task (seen in SKILL.md and references/task-patterns.md).\n
Boundary markers: The documentation notes that arguments must be JSON-serializable, but does not provide explicit prompt delimiters for the data.\n
Capability inventory: The workflow environment allows for network operations (e.g., urllib.request and fetch calls in references/task-patterns.md), task orchestration, and general code execution.\n
Sanitization: The skill lacks instructions for sanitizing or escaping external content before it is processed by the workflow tasks.

render-workflows