The Agent Skills Directory

[SAFE]: The skill operates within the established boundaries of the Renku runtime and CLI, focusing on location and prop authoring. It includes specific instructions to the agent to avoid attempting to bypass permission denials or accessing sensitive system files.
[COMMAND_EXECUTION]: The skill makes legitimate use of the renku CLI to perform project tasks such as reading context, validating operations, and applying design updates. These commands are executed within the local project environment and target files in the tmp/ directory.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface due to its requirement to read external screenplay and asset data to inform production design decisions. 1. Ingestion points: Retrieves external content using renku screenplay show and renku asset list in SKILL.md. 2. Boundary markers: Instructs the agent to resolve identities with the user and avoid automatic data enrichment. 3. Capability inventory: The skill can write operation files to tmp/operations/ and execute state-changing commands like renku location apply and renku prop apply (referenced in SKILL.md, references/location-authoring.md, and references/prop-authoring.md). 4. Sanitization: Implements mandatory --validate and --dry-run checks to ensure all modifications are reviewed before application.

production-designer