delegation
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and automatically update persistent files (
.agents/em-context.mdand.agents/reports/[name].md) based on the ongoing conversation context. This creates a feedback loop where untrusted data or instructions introduced during the conversation could be stored in long-term files and parsed during subsequent executions. - Ingestion points:
.agents/em-context.mdand.agents/reports/[name].mdwithinSKILL.md. - Boundary markers: Missing boundary markers or delimiters when reading or writing the profile content.
- Capability inventory: File system read and automatic write/update capabilities across
SKILL.md. - Sanitization: Absent; the instructions lack rules to filter, validate, or sanitize runtime user inputs before saving them to disk.
Audit Metadata