ai-share-of-voice-audit
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for ingesting untrusted data in the form of answer transcripts provided by the operator for classification and analysis. Ingestion points are located in the multi-model brand mention and citation matrix workflow (SKILL.md) and the transcript classification procedures (references/checks.md). While there are no explicit boundary markers or sanitization steps defined, the skill's capability inventory is strictly limited to analytical reporting and lacks any access to dangerous tools such as network requests, file system modification, or shell command execution, which prevents the content of ingested transcripts from causing system-level harm.
- [DYNAMIC_EXECUTION]: A Python helper function for calculating Share of Voice statistics is included in the reference documentation. This function serves as a static mathematical implementation for the agent to follow during the audit process and does not involve the dynamic execution of untrusted code or the generation of executable strings from external inputs.
Audit Metadata