The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data (the presentation brief) to generate content for slides, which introduces a potential vector for indirect prompt injection.
Ingestion points: The primary ingestion point is the user-provided 'brief' in Phase 0, which the agent analyzes to derive the outline, message, and tonal arc.
Boundary markers: There are no explicit boundary markers or 'ignore' instructions wrapping the processed text before it is used to populate slide shapes.
Capability inventory: The skill possesses extensive write capabilities to the Penpot environment, including board creation, text modification, layout manipulation, and external image uploads via uploadMediaUrl.
Sanitization: No specific sanitization or escaping is performed on the text extracted from the brief before it is passed to penpot.createText calls in the buildSlide.js script.
[COMMAND_EXECUTION]: The skill relies on the execute_code tool to run its internal JavaScript scripts (scripts/*.js) for automating Penpot design tasks.
This behavior is consistent with the primary purpose of the skill and the operational requirements of the Penpot automation platform.
The applyDeckEffects.js script includes a security check to ensure that image uploads only use https:// URLs, reducing the risk of SSRF or local file access through the upload tool.

penpot-build-deck