engineer-motivation
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content through manager context and user profile files, creating an indirect prompt injection attack surface.
- Ingestion points: Reads from
.agents/em-context.mdand.agents/reports/[name].mdto fetch team and engineer motivation profiles (SKILL.md). - Boundary markers: Absent; no explicit delimiters or strict instruction isolation techniques are applied when parsing the profile content.
- Capability inventory: The skill automatically writes or updates files within the
.agents/directory dynamically based on conversation context (SKILL.md). It contains no active network or shell subprocess invocation capabilities. - Sanitization: Absent; content is saved and read directly without filtering, escaping, or strict schema enforcement.
Audit Metadata