The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process user-provided movie scripts and requirements, which constitutes a potential attack surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context through script reading described in SKILL.md and managed by the scripts/design_memory.py utility.
Boundary markers: The skill contains instructions to adhere strictly to the provided script (e.g., in references/production-contract.md), but it lacks formal system-level delimiters to isolate data from instructions.
Capability inventory: The agent is authorized to read and write files within a local project directory using the provided Python script. It does not have network access or arbitrary system execution capabilities.
Sanitization: The Python utility scripts/design_memory.py implements path validation to prevent traversal attacks and strictly validates JSON structures, though it does not sanitize the natural language content of scripts for malicious LLM instructions.
[COMMAND_EXECUTION]: The skill requires the agent to execute a bundled Python script (scripts/design_memory.py) to manage project persistence and design history.
Evidence: SKILL.md and references/design-memory-protocol.md provide operational instructions for the agent to run this script for initialization, context recovery, and data commits.
Safety Review: The script relies solely on the Python standard library. It includes explicit security checks, such as verifying that all file operations remain within the user-defined project_root and forbidding parent directory traversal (..) or alternate data streams.

ai-storyboard-director