The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, including local files and content fetched from GitHub. This creates an indirect prompt injection surface where a malicious file being scanned could attempt to manipulate the scanner's output or behavior. However, the instructions explicitly direct the agent to detect and report such instructions (e.g., 'instructions to ignore security warnings' or 'security bypass'), and the analysis is performed as a static audit of the content.
[EXTERNAL_DOWNLOADS]: The skill fetches repository metadata and file contents from GitHub and GitHub's raw content servers when the --url flag is used. These references target well-known, established services for the purpose of auditing remote code before installation.
[COMMAND_EXECUTION]: Uses basic shell commands like ls and grep to locate and search for patterns within skill directories. These operations are limited in scope and consistent with the tool's auditing purpose.

security-scanner