find-unknowns
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists entirely of instructional markdown and a template. It does not contain any executable code, shell scripts, or external dependencies.
- [INDIRECT_PROMPT_INJECTION]: The skill has a defined attack surface for indirect prompt injection because it is designed to ingest and analyze untrusted data from repositories and external sources.
- Ingestion points: The skill instructs the agent to read "the conversation, named files and artifacts, the repository or data involved" in the 'Inspect before asking' section of SKILL.md.
- Boundary markers: The skill includes strong internal guardrails, stating that "The pass grants no new authority" and that "Silence or 'I don't know' is never consent to a consequential or irreversible step."
- Capability inventory: The skill uses inspection and search capabilities to gather information but requires explicit user authorization for any new external actions or resource-intensive tasks.
- Sanitization: While no specific sanitization logic is provided in the instructions, the skill mandates a 'batched' question approach where findings are presented to the user for verification before any action is taken.
Audit Metadata