The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from various local files to adapt image templates, which introduces a vulnerability surface if those files contain untrusted content.
Ingestion points: The skill reads from squads/{code}/_build/discovery.yaml, _opensquad/_memory/company.md, _opensquad/_memory/preferences.md, and other discovery/design files.
Boundary markers: The instructions do not specify any delimiters or safety markers to differentiate between the data being read and potential instructions embedded within that data.
Capability inventory: The agent can write HTML and Markdown files to the squad's directory and invokes the image-creator skill to perform rendering tasks.
Sanitization: There is no evidence of data sanitization or validation performed on the context files before they are used to generate HTML code.
[DYNAMIC_EXECUTION]: The skill creates self-contained HTML/CSS files at runtime by interpolating user-provided context into base templates. These files are then dynamically processed by the rendering tool. While this is a core requirement for a design agent, it represents the generation and execution of code at runtime.

template-designer