scrub
Pass
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted input from conversation history to identify 'wrong claims' and uses these to perform targeted searches across project files. This creates a surface for indirect prompt injection where a malicious actor could influence search parameters or report content.
- Ingestion points: Recent conversation context and project files (INVARIANTS.md, AGENTS.md, docs, memory files, source code).
- Boundary markers: Phase 1 includes a mandatory confirmation step where the agent must wait for the user to approve the 'Suspected hallucination' and 'Correct fact' before searching.
- Capability inventory: The skill utilizes file read capabilities (grep) and file write capabilities (creating reports in the 'scrub-reports/' directory).
- Sanitization: No specific sanitization or escaping of user-provided 'key terms' is defined for the grep operations.
Audit Metadata