The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing the prism CLI tool to manage platform state. It includes specific safety protocols, such as verifying the active organization, requiring user confirmation before state-changing operations (like deployment or activation), and using explicit placeholders for IDs to prevent accidental execution of commands.
[EXTERNAL_DOWNLOADS]: The skill references the @prismatic-io/prism npm package for environment setup. It instructs the agent to check the existing toolchain first and seek explicit user approval before performing any installations or upgrades. It also fetches product documentation from the official domain at prismatic.io/docs/llms.txt.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources including platform logs, execution results, and documentation. It mitigates potential risks by instructing the agent to distinguish observed state from inference, verify targets across complete result sets, and validate all commands against their current contracts before use.
[DATA_EXFILTRATION]: The instructions contain protective measures, explicitly forbidding the agent from asking users to paste secrets into the conversation or displaying secret values returned by tools, directing them instead to secure configuration surfaces.

orby