ai-prune
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill parses the content of various repository files, including logs and documentation, to determine which items are outdated. Malicious text within these files could potentially trick the agent into trashing important data or failing to prune sensitive leftovers.
- Ingestion points: Files in
.ai/,.cache/,README.md,AGENTS.md,CLAUDE.md, andgit log. - Boundary markers: Absent; the agent reads files directly without wrapping content in protective delimiters.
- Capability inventory: The skill can archive files (
mkdir,mv) and delete them via the macOS Trash utility (/usr/bin/trash). - Sanitization: None; the agent processes raw file content.
- [COMMAND_EXECUTION]: The skill relies on executing shell commands like
fd,git, and/usr/bin/trash. While appropriate for a maintenance tool, this exposes a standard surface for command-related issues if inputs were not handled correctly.
Audit Metadata