The Agent Skills Directory

[SAFE]: The skill performs local file operations within the workspace to build a landing page. Analysis of the Python scripts shows no network requests, subprocess calls, or access to sensitive system directories.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from local files like copy.md and strategy.json to generate an HTML page.
Ingestion points: Data is read from multiple files in workspace/pages/ and workspace/brand/ by scripts/build-page.py.
Boundary markers: Not explicitly defined in the source data files.
Capability inventory: The skill writes output files (index.html, meta.json, qa.md) to the local workspace.
Sanitization: The build script utilizes html.escape() when interpolating text content into the HTML template, mitigating the risk of content-based injections.

page-build