kai-brand-pulse
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes local Python modules and scripts, such as
kai.source_data.collectandscripts/intel/brand_pulse.py, to perform data aggregation and analysis tasks. These scripts are described as internal components of the agent framework. - [PERSISTENCE]: The skill provides documentation for setting up a weekly cron job to automate brand monitoring. While this is a documented feature for the user to implement, it represents a standard persistence mechanism for recurring execution.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data collected from various public platforms, including Reddit, X, LinkedIn, and general web search results, which are then used to generate marketing analysis and copy.
- Ingestion points: Data packets located in
platforms/*.mdcontaining content fetched from external social media and news sources. - Boundary markers: The skill requires the use of citation IDs and source IDs for all generated claims to ensure evidence-based analysis.
- Capability inventory: Subprocess execution of Python scripts and file system writes to the
workspace/directory. - Sanitization: The instructions do not detail specific automated sanitization logic for the ingested external text, relying instead on manual citation and data-gap reporting workflows.
Audit Metadata