The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates the ingestion of data from arbitrary external websites, which presents a risk of Indirect Prompt Injection.
Ingestion points: Web content is ingested into the agent context via the scrapling extract command group (SKILL.md) and the MCP tools such as get, fetch, and stealthy_fetch (references/mcp-server.md).
Boundary markers: The instructions do not recommend or implement delimiters (such as XML tags or unique markers) to isolate ingested web content from agent instructions.
Capability inventory: The skill has the capability to perform network requests (all HTTP methods), write content to local files (scrapling extract ... output.md), and execute shell commands via the CLI.
Sanitization: While the skill can convert HTML to Markdown or plain text, it does not include filters or validation to detect and neutralize instructions embedded in the scraped text content.
[EXTERNAL_DOWNLOADS]: The skill's setup instructions involve downloading the scrapling library from the Python Package Index (PyPI) and browser drivers via the scrapling install command. These are standard operations for the tool's core functionality.
[COMMAND_EXECUTION]: The agent is instructed to execute shell commands to install the framework and use its command-line interface for scraping and content extraction tasks.
[DATA_EXFILTRATION]: The skill provides tools for performing various HTTP methods, including POST and PUT, which allows for data to be transmitted to external domains. While intended for legitimate interactions, this capability can be leveraged for data transmission.

scrapling