The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill explicitly instructs the agent to run commands outside the security sandbox using dangerouslyDisableSandbox: true. This is requested to allow the network and file system access necessary for web scraping, but it bypasses standard security boundaries.
[EXTERNAL_DOWNLOADS]: The skill directs the user to install the slurp-ai package globally via npm install -g slurp-ai. This introduces an external dependency that executes with full user permissions.
[COMMAND_EXECUTION]: The skill relies on executing the slurp binary and a local script analyze-sitemap.js to perform its core functions of network crawling and file compilation.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest large amounts of documentation from external websites into the agent's context. This creates a significant surface for indirect prompt injection.
Ingestion points: Content is fetched from arbitrary URLs via the slurp command and stored in slurp_compiled/compiled_docs.md.
Boundary markers: None are specified in the skill to delimit the scraped content or warn the agent about potentially malicious instructions within the documentation.
Capability inventory: The skill uses slurp for network requests and file writing, and node to execute the analyze-sitemap.js utility script.
Sanitization: The skill does not mention any sanitization or filtering of the scraped markdown content before it is provided to the agent.

documentation-scraper