pluck
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill interacts with a local Rust daemon (
pluckd) and wraps theripgrepCLI tool for search functionality. Theinitcommand with the--mode aggressiveflag suggests initialization routines that modify configuration files for agents like Claude Code or Cursor. - [EXTERNAL_DOWNLOADS]: The skill performs automatic downloads of Tree-sitter grammar files from external repositories during its first run to enable language-specific parsing features.
- [PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted code from local repositories into the agent's context, which is an indirect prompt injection surface. \n
- Ingestion points: The
read,search, andsymboltools inSKILL.mdextract arbitrary text from the filesystem.\n - Boundary markers: No explicit security delimiters are defined to prevent the agent from executing instructions embedded in the retrieved code.\n
- Capability inventory: The skill has the ability to read all files in a repository and analyze call graphs via the
pluckddaemon.\n - Sanitization: There is no evidence of sanitization or filtering for the content retrieved from the codebase.
Audit Metadata