leadership-development-diagnostic
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze natural language data from external stakeholders, including 360-degree feedback, email samples, and meeting transcripts, which creates a surface for indirect prompt injection attacks. The instructions lack explicit boundary markers or directions for the agent to treat this external content as untrusted data rather than instructions.
- Ingestion points: As specified in references/evidence-and-analysis.md, the skill processes 360-degree feedback from multiple rater groups, direct meeting observations, and work products such as emails.
- Boundary markers: The prompt templates and sequences defined in the skill do not utilize delimiters or specific instructions to isolate external data from the agent's core instructions.
- Capability inventory: The agent uses ingested data to produce formal readiness verdicts and behavior change plans. While it does not have direct access to system commands or network operations, the injection could compromise the integrity of the diagnostic output.
- Sanitization: There are no requirements for sanitizing or filtering the external content for potential injection patterns.
Audit Metadata