mole
Pass
Audited by Gen Agent Trust Hub on Nov 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its interaction with untrusted system data. \n
- Ingestion points: The agent reads file paths, directory entries, and system status via
mo analyze --json,mo status --json, and the candidate list inclean-list.txt(SKILL.md). \n - Boundary markers: There are no explicit instructions to use delimiters or warnings to ignore embedded instructions in the processed system output. \n
- Capability inventory: The skill can execute file removals and app uninstalls via
mo clean,mo purge, andmo uninstall(references/macos.md, references/windows.md). \n - Sanitization: No sanitization or escaping of external content is specified before interpolation into agent responses. \n- [COMMAND_EXECUTION]: The skill uses the
moCLI for its primary functions. It mitigates execution risks by requiring--dry-runpreviews for all destructive tasks and mandating explicit user consent for actual removals. It specifically forbids the use ofsudo, administrative elevation, and direct shell commands likermorfind -delete, effectively confining the agent to the CLI's own safety logic. \n- [EXTERNAL_DOWNLOADS]: The skill references installation resources from the author's site (mole.fit) and official GitHub repositories (github.com/tw93/Mole). These are documented as trusted or well-known sources. The skill adheres to safe practices by instructing the agent to provide these links to the user for manual execution rather than downloading or piping remote scripts directly into a shell.
Audit Metadata