The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Instructions describe installing official tools using piped shell scripts from vendor and trusted domains (e.g., https://cli.runpod.net | bash and https://astral.sh/uv/install.sh | sh).
[COMMAND_EXECUTION]: The skill provides extensive guidance on executing commands on remote infrastructure via SSH, including managing long-running processes and non-interactive environment setup.
[CREDENTIALS_UNSAFE]: Mentions the management of sensitive credentials such as RUNPOD_API_KEY and S3 keys. The guidance follows best practices by recommending environment variables and standard configuration files for storage.
[PROMPT_INJECTION]: The documentation defines handler interfaces for serverless workers that process untrusted data from job["input"].
Ingestion points: reference/docker.md (job processing logic)
Boundary markers: Not explicitly defined in examples
Capability inventory: reference/pod-workflows.md (SSH and shell command execution)
Sanitization: Not explicitly covered in the reference implementation
[NO_CODE]: The skill is entirely instructional and does not contain any executable logic or automated scripts that run in the agent's environment.

runpod-usage