The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/run.py uses os.execv to perform process replacement. This allows the skill to re-execute itself using a Python interpreter found in a local virtual environment (.runners-venv). This is a common and safe bootstrap pattern used in this agent framework to manage dependencies without manual user intervention.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text through parameters like --top, --bottom, and --context, which are then interpolated into prompts for image generation models. This constitutes a standard injection surface for untrusted data. However, the risk is assessed as safe because the capabilities of the skill are restricted to image generation and file system operations within a dedicated ./generated/ directory, and it does not allow the execution of instructions extracted from the user text by the agent itself.

meme-card-maker