The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted user-provided articles and interpolates them into a generated HTML preview file (index.html) without explicit instructions for sanitization or escaping. This creates a surface for indirect prompt injection or cross-site scripting (XSS) within the local preview environment. 1. Ingestion points: Article content is collected directly from user input in Step 1 of the SKILL.md workflow. 2. Boundary markers: The skill lacks explicit instructions to wrap the article content in delimiters or include warnings to ignore embedded instructions. 3. Capability inventory: The agent has the capability to write files to the local project directory and execute shell commands for building or starting servers as described in Step 7. 4. Sanitization: No sanitization or HTML escaping logic is defined for the interpolated content before it is placed in the template, although the skill advises the agent to avoid scripts in the final copied payload.

space-wechat-layout