The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and analyzes HTML content from external websites provided by the user, which is a standard surface for indirect prompt injection.
Ingestion points: The skill uses the webfetch tool to retrieve HTML from arbitrary user-provided URLs (SKILL.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the fetched HTML are present.
Capability inventory: The skill executes local Python scripts (dfs_client.py, report_publish.py) and uses the webfetch tool.
Sanitization: There is no evidence of HTML sanitization or filtering before the content is processed by the agent.
[COMMAND_EXECUTION]: The skill relies on executing local Python scripts to perform audits and generate client reports.
Evidence: The instructions call for running python scripts/dfs_client.py with user-supplied URLs and python scripts/report_publish.py to generate the final HTML/PDF deliverables.

image-seo