how
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill interpolates external data, including user questions and subagent findings, into its reasoning prompts.
- Ingestion points: The
{QUESTION}and{EXPLORER_FINDINGS_ALL}placeholders in the reference prompt templates. - Boundary markers: Content is delimited using standard Markdown blockquotes rather than strict structural separators like XML tags.
- Capability inventory: The tools available to subagents are limited to read-only filesystem inspection (
Read,Grep,Glob). - Sanitization: No automated validation or filtering is applied to the data before interpolation into the subagent context.
- [SAFE]: The skill consistently applies
readonly: trueto all spawned subagents, which prevents unauthorized filesystem modifications or persistent changes to the environment.
Audit Metadata