The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The file references/logo-creator.md suggests using Python code involving subprocess and cairosvg to convert SVG logos. This pattern encourages the agent to generate and execute code at runtime.
[COMMAND_EXECUTION]: The SKILL.md file mentions a Python hook script located at hooks/brand_guard.py for automated branding enforcement. Because this file is not provided in the skill package, its behavior remains unverifiable, constituting a potential security risk.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing external data and user-controlled configuration.
Ingestion points: The agent reads from brand-config.md (populated by user answers) and documents from /mnt/user-data/uploads/.
Boundary markers: None identified. Instructions do not mandate the use of delimiters or 'ignore' instructions for data interpolation.
Capability inventory: The skill allows for Python code execution (suggested for logos), file system writing (brand-config.md), and automated document generation (Word, PPTX, HTML).
Sanitization: None identified. Data provided by the user is directly interpolated into various output formats without validation.

brand-studio