modifying-taxonomic-filter
Pass
Audited by Gen Agent Trust Hub on Jun 24, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill outlines strict engineering practices for modifying a complex UI component, emphasizing architectural consistency and the preservation of telemetry contracts.
- [COMMAND_EXECUTION]: The skill references the use of an internal testing utility,
hogli, which is used for running local tests within the project's frontend directory. This is standard developer workflow and does not involve unauthorized command execution. - [DATA_EXPOSURE]: Includes SQL query templates for refreshing 'Product Reality' data. These queries are designed to be run against the organization's own PostHog instance to gather aggregated usage statistics (ratios) and do not facilitate the exposure of sensitive user data to external parties.
- [INDIRECT_PROMPT_INJECTION]: While the skill instructs the agent to make decisions based on external telemetry data (Product Reality section), the instructions emphasize that critical changes like ordering or promotion require explicit human sign-off, mitigating risks associated with automated decision-making based on potentially poisoned data.
Audit Metadata