The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill features a surface for indirect prompt injection as it ingests and processes untrusted data from screenshots and codebase files.
Ingestion points: Visual analysis of screenshots and extraction of color definitions from style sheets and component files (SKILL.md).
Boundary markers: The skill does not define specific markers or instructions to isolate data from commands during ingestion.
Capability inventory: The skill utilizes toolsets for file reading, writing, editing, and web fetching (SKILL.md frontmatter).
Sanitization: There is no evidence of content sanitization or validation for the data extracted from users' project files.
[COMMAND_EXECUTION]: The documentation includes instructions for executing shell commands to perform automated accessibility audits.
Evidence: Use of npx lighthouse and npx pa11y for scanning URLs (SKILL.md).
[EXTERNAL_DOWNLOADS]: The skill references and suggests the use of well-known external accessibility services and tools.
Evidence: Recommends WebAIM, Coolors, and Adobe Color tools for contrast validation (SKILL.md).

color-contrast-auditor