The Agent Skills Directory

[DATA_EXFILTRATION]: The skill reads API credentials from ~/.gooseworks/credentials.json. This information is then used in Authorization headers for network requests to the vendor's API endpoint at api.gooseworks.ai.
[COMMAND_EXECUTION]: The setup instructions use python3 -c shell commands to extract data from the local JSON credentials file and curl to perform network operations.
[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection by ingesting and processing untrusted data from arbitrary web pages.
Ingestion points: Web content is ingested through the website_url and url parameters in the SKILL.md file.
Boundary markers: The instructions lack delimiters or explicit safety warnings to prevent the agent from obeying malicious instructions potentially embedded in the scraped content.
Capability inventory: The skill has network access via curl and executes shell-based commands for setup.
Sanitization: No validation or sanitization of the content retrieved from external websites is performed before it is passed to the agent's context.

ai-web-scraping-scrapegraph