The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from application logs and possesses a high-privilege capability inventory, creating a surface for indirect prompt injection attacks.
Ingestion points: The skill uses list_logs() to retrieve build, runtime, and HTTP error logs from Render services which are controlled by application behavior.
Boundary markers: The instructions do not define boundary markers (e.g., XML tags or delimiters) or explicit 'ignore embedded instructions' warnings for the agent when analyzing log content.
Capability inventory: The skill has access to powerful tools including query_render_postgres (arbitrary SQL execution), update_environment_variables (service configuration modification), and shell-based CLI commands for deployment management.
Sanitization: No sanitization or filtering of log data is performed before the content is interpreted by the agent.
[COMMAND_EXECUTION]: The skill utilizes the render CLI and suggests local filesystem operations (editing source code) and version control actions (git commit/push). While these are intended for debugging and fixing issues, they provide the agent with broad execution capabilities within the user's environment.
[SAFE]: The skill follows secure practices for credential management by instructing the user to configure API keys via the official Render dashboard and environment variables rather than hardcoding secrets.

render-debug