kelly-beauty-intel
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a robust 'review-first' workflow where all signals, actions, and drafts generated by the agent must be reviewed and approved by a human operator in the provided AirApp UI before any further action is taken.
- [DATA_EXPOSURE]: No hardcoded credentials were found. The skill correctly instructs users not to share API keys in chat and uses environment variables for local script execution, following security best practices for industry-standard tooling.
- [COMMAND_EXECUTION]: Local scripts provided (setup.mjs, publish_airapp.mjs) are transparent utility tools for managing the Busabase workspace and AirApp deployment. They do not exhibit suspicious patterns or download/execute untrusted remote code.
- [OBFUSCATION]: Analysis of the bundled vendor files (e.g., busabase-airapp-gate.js) and application logic confirmed they are standard minified or bundled assets rather than malicious obfuscation. No hidden commands, zero-width characters, or homoglyph attacks were detected.
- [INDIRECT_PROMPT_INJECTION]: While the skill ingests external data (competitor offers, news, trends), it incorporates strong mitigation by design. The instructions prioritize 'blocked claims' for medical/clinical content and mandate human approval, effectively neutralizing the risk of the agent obeying instructions embedded in external signals.
- [DYNAMIC_EXECUTION]: The use of esbuild-wasm and dynamic imports is limited to standard build-time bundling and conditional module loading for the application's UI providers (demo vs. live), posing no security risk.
Audit Metadata