The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes several diagnostic scripts (analyze-styles.ts, extract-tokens.ts, analyze-accessibility.ts) that ingest and process local CSS, HTML, and component source files. Because these files may contain content from untrusted third parties (e.g., external libraries or legacy codebases), they represent an attack surface for indirect prompt injection where malicious instructions could be embedded in comments or strings to influence the agent's behavior.
Ingestion points: Untrusted data enters the agent context through Deno.readTextFile operations in scripts that audit CSS, extract design tokens, and perform accessibility checks.
Boundary markers: The scripts do not use explicit delimiters or specific 'ignore' instructions when presenting parsed content or analysis summaries to the agent.
Capability inventory: The skill requests Deno --allow-read and --allow-write permissions, enabling the agent to inspect the local filesystem and generate new source code files.
Sanitization: While the scripts use regular expressions to extract specific design-related data (colors, spacing, typography), they do not perform security-focused sanitization of the broader content before presenting the analysis output to the agent.

frontend-design