sentiment-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent on how to build systems that ingest and analyze untrusted text from external sales prospects.
- Ingestion points: The skill explicitly deals with processing external "prospect messages" as described in SKILL.md.
- Boundary markers: The provided LLM prompt example uses basic double quotes to delimit untrusted input (
Message: "[prospect message]") but does not implement robust boundary markers or warnings to ignore embedded instructions. - Capability inventory: The skill is strictly documentation-based and does not include executable scripts, network capabilities, or file system access.
- Sanitization: The guidance does not include specific technical recommendations for sanitizing or escaping the prospect's input before it is processed by a sentiment model or LLM.
Audit Metadata