marketing-attribution-dashboard
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines logic for processing untrusted data from URL parameters (UTMs) and HTTP referrers to classify marketing channels. This is a standard data ingestion surface for analytics tasks and is handled via logic-based classification.
- Ingestion points:
SKILL.md(viaclassifyChannelfunction). - Boundary markers: Not explicitly implemented in the logic snippets.
- Capability inventory: The skill's logic is limited to classification and mathematical weighting; it does not utilize sensitive capabilities like arbitrary file writing or network exfiltration.
- Sanitization: Logic employs substring and regex matching for identification.
- [COMMAND_EXECUTION]: The evaluation tasks include instructions for the agent to run verification scripts using standard development tools.
- Evidence: Use of
npx ts-nodeto execute test scripts in theevals/directory to verify the correctness of the implementation.
Audit Metadata