The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data from reports and workspace files to generate executable code such as HTML and JavaScript. This creates a surface for indirect prompt injection where malicious content within a report could influence sub-agents to generate harmful scripts or UI components.
Ingestion points: Reads report content from reports/*.md and _workspace/*.md.
Boundary markers: No explicit delimiters or instructions are provided to the sub-agents to ignore embedded instructions within the source documents.
Capability inventory: The sub-agents have the capability to write files to the web/ directory, including index.html and multiple .js files.
Sanitization: There are no defined steps for sanitizing or validating the content extracted from the reports before it is interpolated into the generated web assets.
[COMMAND_EXECUTION]: The skill provides instructions for the user to execute shell commands to preview the generated website, such as python3 -m http.server 8000 or npx serve .. While these are standard development practices, they involve running a local web server that exposes the generated files to the local network.

report-to-webpage