The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources such as screenplay scenes, character profiles, and user-provided JSON configuration files. This creates a surface for indirect prompt injection where instructions could be embedded in character descriptions or design notes. \n
Ingestion points: Operations like renku cast design show, renku cast context, and loading files via the --file argument. \n
Boundary markers: Instructions explicitly require using renku ... validate commands before applying any changes, ensuring schema compliance. \n
Capability inventory: The skill can write files (renku cast design write), update project states (renku cast apply), and perform network operations (renku cast voice attach). \n
Sanitization: Validation is performed via CLI tools to ensure data conforms to the expected JSON contracts described in the reference documentation. \n- [COMMAND_EXECUTION]: The skill utilizes the renku CLI to manage project metadata and character facts. This is the intended primary purpose of the skill and follows standard development patterns for CLI-integrated agents. \n- [EXTERNAL_DOWNLOADS]: The skill facilitates communication with ElevenLabs (a well-known service) to fetch and attach voice samples. The instructions specify that the agent must ask for user approval before executing network-active commands that interact with external providers.

casting-director