The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read local configuration files based on user-provided brand names, creating a potential path traversal surface.\n- Ingestion points: User-specified brand names in image requests (e.g., 'Digital Samba' in examples.md).\n- Boundary markers: Not specified in the instruction set.\n- Capability inventory: Reading from the brands/ directory and execution of tools/ideogram4.py which interacts with an external API (found in SKILL.md).\n- Sanitization: No evidence of input validation or path sanitization.\n- [COMMAND_EXECUTION]: Uses a local utility script tools/ideogram4.py to communicate with the Ideogram API via shell commands. This is a primary function of the skill.

ideogram4