The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables the agent to fetch and process data from untrusted external websites, creating a surface for embedded instructions to influence agent behavior.
Ingestion Point: Content is retrieved from remote URLs using the crawlberg scrape command and stored in variables such as markdown.content.
Capability Inventory: The agent is given access to execute shell commands and perform network requests via the scraping tool.
Boundary Markers: The instructions do not prescribe the use of boundary markers or system instructions to ignore commands within the fetched data.
Sanitization: No sanitization or filtering logic is provided to handle potentially malicious instructions in the scraped content.
[COMMAND_EXECUTION]: The skill provides numerous examples for executing the crawlberg command-line utility.
Evidence: Commands such as crawlberg scrape https://spa.example.com --browser-mode always and WebSocket connections via --browser-endpoint are documented for the agent to use.

headless-fallback