The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads project-specific documentation and configuration files to build a 'brief' which is then used to direct the actions of a sub-agent or CLI tool.
Ingestion points: Reads DESIGN.md, CONTEXT.md, and README files from the user's repository in Step 1 to extract scene rules.
Boundary markers: No explicit delimiters or 'ignore embedded instructions' warnings are specified for the generated brief.
Capability inventory: Performs shell command execution (codex exec), background process management (PID tracking), file system operations (symlinking node_modules, creating copies), and browser interaction (opening live pages).
Sanitization: There is no mention of validating or sanitizing the content of the project files before they are included in the execution brief.
[COMMAND_EXECUTION]: The skill relies on shell commands for its core functionality, including environment setup and task execution.
Evidence: Uses codex exec for model generation, ln -s for symlinking directories, and process management via PID tracking in Step 4.
[DYNAMIC_EXECUTION]: The skill dynamically generates a markdown 'brief' at runtime which acts as the instruction set for a sub-execution environment.
Evidence: Step 5 pipes a dynamically created brief.md into the codex exec command.
[EXTERNAL_DOWNLOADS]: The instructions encourage the agent to update software packages or fetch newer versions of CLI tools if compatibility issues arise.
Evidence: Step 5 instructs the agent to 'take a newer one... or update the package' if the CLI tool is outdated.

make-3d