The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in scripts/ai_label.py to call exiftool for writing IPTC metadata to images. This execution is performed using a list of arguments and does not invoke a shell, mitigating command injection risks.
[COMMAND_EXECUTION]: The SKILL.md instructions guide the agent to execute chrome --headless for PDF generation and higgsfield for image creation. These are standard operations for the intended use case of publishing branded assets.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from a file named BRAND.md to determine design constraints and text patterns.
Ingestion points: BRAND.md file content is read and interpreted by the agent (mentioned in SKILL.md).
Boundary markers: None. The agent is instructed to follow BRAND.md as the primary authority.
Capability inventory: The skill can execute shell commands via subprocess (in scripts/ai_label.py), perform network requests (implied by image generation tools), and write files to the local system.
Sanitization: No explicit sanitization or filtering of the instructions found within BRAND.md is specified, creating a surface where a malicious brand kit could influence agent behavior.

brandkit-publish