The Agent Skills Directory

[DYNAMIC_EXECUTION]: The entry point scripts/run.py uses os.execv to ensure the skill runs within its designated virtual environment. The path for the Python interpreter is resolved relative to the script's installation directory. This is a standard mechanism for environment isolation.
[INDIRECT_PROMPT_INJECTION]: The skill generates image prompts based on user-supplied headlines and CTAs. This creates a surface for indirect prompt injection, characterized by the following:
Ingestion points: Untrusted text enters via command-line arguments as described in SKILL.md and is passed to the generation logic in scripts/run.py.
Boundary markers: Instructions in SKILL.md specify that user-provided content is encapsulated in double quotes during prompt construction.
Capability inventory: The skill uses tools for file manipulation (Read, Write, Edit, Grep, Glob) and shell execution (Bash). Its primary function involves creating image files and invoking image models through the platform batch runner.
Sanitization: Content length is restricted by design constraints (e.g., headlines limited to 7 words), which limits the complexity of potential injection payloads.

banner-maker