The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches command specifications and security research from gtfobins.org and downloads test suites from the official GNU coreutils and uutils GitHub repositories. These are well-known and reliable sources for software development and security auditing.
[REMOTE_CODE_EXECUTION]: Content is downloaded from trusted GitHub organizations and extracted locally using curl | tar. This is a standard procedure for obtaining reference test suites and, given the trusted nature of the sources, does not pose a security risk in this context.
[COMMAND_EXECUTION]: The agent is instructed to run go test and related shell commands to verify the builtin implementation, perform fuzzing, and conduct exploratory testing within the development environment.
[PROMPT_INJECTION]: The skill explicitly recognizes the threat of indirect prompt injection from processing external web content. It implements proactive mitigations by instructing the agent to treat all external data as untrusted, use specific delimiters for external content, and ignore any instructions or overrides found within that data.

implement-posix-command