The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The dashboard_designer.py script reads data from an external JSON file and incorporates it into a Markdown-formatted report, creating a surface where malicious input could influence an agent processing the output.
Ingestion points: Untrusted data is ingested from the file path provided to the --input argument of the dashboard_designer.py script as described in the SKILL.md workflow.
Boundary markers: The generated audit report does not utilize boundary markers or explicit instructions to the AI agent to isolate or ignore embedded instructions within the processed content.
Capability inventory: The skill includes a Python script capable of reading local files and writing to the file system if the --output flag is used.
Sanitization: The script performs no validation, escaping, or sanitization of the strings retrieved from the input JSON before they are rendered into the final Markdown output.

metrics-dashboard