stelow-product-evolutionary-principles
Pass
Audited by Gen Agent Trust Hub on Jul 18, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The subagent documentation (subagents.md) strictly enforces the use of 'fresh context' for all delegated tasks. This implementation pattern is a security best practice that ensures sub-tasks are isolated, preventing potential instruction leakage, context rot, or accidental adherence to previous session data.
- [EXTERNAL_DOWNLOADS]: The skill provides instructions for the agent to install or use several external tools and extensions from the Pi/Stelow ecosystem, such as @tintinweb/pi-subagents, @earendil-works/pi-agent-browser, and @plannotator/pi-extension. These are documented as standard functional dependencies for browsing, visual review, and task delegation within the orchestrator's framework.
- [COMMAND_EXECUTION]: The orchestrator instructions include shell command patterns for project management tasks, such as atomic file locking (file-locking.md), reading JSON configuration via Node.js (read-config.md), and executing project-specific test suites (execution-loop.md). These operations are standard for development automation and are clearly scoped to the local project environment.
- [DATA_EXFILTRATION]: While the skill includes browsing capabilities via the agent-browser tool, its use is specifically directed toward rendered UI verification, accessibility audits, and interaction testing on project-related URLs (e.g., localhost). There are no indications of unauthorized data transmission or suspicious network destinations.
Audit Metadata