The Agent Skills Directory

[PROMPT_INJECTION]: The SKILL.md file contains a 'Reporting' section that provides specific behavioral instructions for the agent, such as how to format findings and when to 'Block' or 'Approve' based on its analysis. While these are instructions targeting agent behavior, they are specific to the skill's intended purpose as a code auditor.
[INDIRECT_PROMPT_INJECTION]: As a tool designed to analyze and process user-provided code and style files, this skill possesses an attack surface for indirect prompt injection. Malicious instructions could theoretically be embedded within the files being audited to influence the agent's results.
Ingestion points: User-provided source files and CSS styles.
Boundary markers: None specified in the skill instructions.
Capability inventory: The skill defines audit rules but does not include capabilities for external network requests or sensitive file writes.
Sanitization: No sanitization of the input code is defined.
[DYNAMIC_EXECUTION]: The recipes.md file includes a setTheme utility that dynamically generates and injects a CSS <style> block into the document head at runtime to temporarily disable transitions. This is a standard and safe implementation for theme switching in web applications.

defaults