The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install the squirrel CLI tool using dangerous piped execution patterns: curl -fsSL https://squirrelscan.com/install | bash and irm https://squirrelscan.com/install.ps1 | iex. These commands download and run arbitrary code from the internet with the current user's privileges without any integrity verification.
[EXTERNAL_DOWNLOADS]: The skill downloads binaries and installation scripts from squirrelscan.com, which is an external source.
[COMMAND_EXECUTION]: The skill's core functionality involves executing the squirrel CLI tool and various shell commands to perform website audits and modify local project files.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface (Category 8) because it crawls and processes data from external websites. 1. Ingestion points: Content from external websites crawled during the audit and crawl processes. 2. Boundary markers: No delimiters or protective instructions are identified to prevent the agent from following commands embedded in crawled data. 3. Capability inventory: Execution of the squirrel CLI, the ability to write to the local file system to apply fixes, and the spawning of subagents via the Task tool. 4. Sanitization: No evidence of sanitization or filtering of external content before it is processed by the agent.

audit-website