money-diagnose
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill documentation includes a benign bash command snippet intended for manual user configuration as a process guardrail.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user business data in Phase 1 (SKILL.md). It employs mandatory boundary markers (a four-phase workflow and explicit human confirmation gate) that isolate untrusted input from action recommendations. Capability inventory is limited to writing a local lock file (~/.smtm/.diagnose-pending), and sanitization is achieved through precise logical deconstruction of input terms.
- [SAFE]: Writing a state file to a specific hidden path is a transparent mechanism for workflow management and does not target sensitive data or facilitate malicious persistence.
Audit Metadata