time-to-close-prediction
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external conversation data to extract features like sentiment and buying signals, which creates a vulnerability surface for indirect prompt injection where malicious instructions in the conversation could influence the agent's logic.
- Ingestion points: Untrusted conversation text is retrieved using the
get_conversations(deal)function within thepredict_from_conversationlogic inSKILL.md. - Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings for the data being analyzed.
- Capability inventory: The skill's pseudo-code is limited to analytical and mathematical operations (e.g., calculating response times, stakeholder counts, and sentiment trends) and does not contain subprocess calls, network operations, or file system writes.
- Sanitization: No filtering or sanitization of the external conversation content is described in the implementation guidance.
Audit Metadata