The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data including 'brand/design-system sources' and 'design tasks'. This creates an attack surface where malicious instructions could be embedded in these inputs to influence the agent's behavior during theme contract generation or downstream styling tasks. Ingestion points: External brand sources, design system documentation, and task descriptions mentioned in the processing steps in SKILL.md. Boundary markers: The instructions explicitly warn against silently replacing approved sources and require separate approval records for changes, which provides a procedural safety boundary. Capability inventory: The skill directs the agent to 'Write or reference the contract in the canonical design package', which involves file system operations to store the theme specification. Sanitization: The skill lacks explicit instructions for sanitizing or escaping content from external sources before including them in the final theme contract output.
[NO_CODE]: The skill does not include any executable scripts, binary files, or installation manifests. It relies purely on natural language instructions and static configuration files, which significantly limits the risk of direct malicious code execution or privilege escalation.

ui-design-theme