new-hire-detection
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Sales Navigator and LinkedIn and interpolates it into prompts for the qualifier agent, which could be exploited if the source data contains malicious instructions.
- Ingestion points: Data from the Sales Navigator model and LinkedIn company enrichment are read in
infra/plays/route-new-hires.tsandreferences/crm-adaptation.md. - Boundary markers: The prompt template for the
icpQualifieragent lacks clear delimiters or 'ignore' instructions for the interpolated company object and lead details. - Capability inventory: The workflow includes capabilities to post to Slack (
slack.postMessage) and perform write operations in CRMs (crm.upsertRecords,crm.insertRecord,crm.createAssociation). - Sanitization: Input data is passed as a JSON string to the LLM without filtering or validation of the content against injection patterns.
Audit Metadata