The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from external websites, which serves as a potential vector for indirect instructions.
Ingestion points: The scripts/scrape.py script extracts information such as page titles, image URLs, and alt text from user-supplied URLs.
Boundary markers: Scraped content is interpolated directly into agent prompts and tool arguments without the use of clear delimiters or explicit instructions for the AI to ignore embedded commands.
Capability inventory: The skill environment allows for downloading files via curl, writing local data with the Write tool, and executing the astria CLI for generation and model management.
Sanitization: While the scripts perform basic text trimming and filtering for known non-image assets, they do not sanitize text for potential prompt injection payloads.
[COMMAND_EXECUTION]: The scripts/templatize.py script programmatically executes the astria CLI tool using the subprocess module.
It uses argument lists for subprocess.run(), which is a recommended practice to avoid shell injection vulnerabilities.
The command arguments are dynamically constructed using data derived from the scraping process and user-provided inputs.

templatize-page