The Agent Skills Directory

[COMMAND_EXECUTION]: Several files, including references/copywriting-formulas.md, references/layout-patterns.md, and references/slide-strategies.md, contain instructions for the agent to execute a Python script located at .claude/skills/design-system/scripts/search-slides.py. This script is not included in the provided skill files. Executing scripts from paths outside the skill directory or using missing dependencies can lead to unexpected behavior if a malicious file is placed at that location.
[EXTERNAL_DOWNLOADS]: The references/html-template.md file fetches the Chart.js library from the official jsDelivr CDN. This is a well-known technology service used for data visualization.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes untrusted data to generate executable HTML/JS content.
Ingestion points: User input enters the skill through the $ARGUMENTS variable in SKILL.md and is processed by the task in references/create.md.
Boundary markers: The skill lacks delimiters or explicit instructions for the agent to ignore potentially malicious commands embedded within the user's presentation content.
Capability inventory: The skill generates HTML, CSS, and Chart.js code for browser execution. While the provided scripts do not perform file writes or network calls, the generated content is rendered in the user's context.
Sanitization: No input sanitization or HTML escaping is visible before the user data is interpolated into the slide templates.

slides