skills/heliohq/ship/e2e/Gen Agent Trust Hub

e2e

Pass

Audited by Gen Agent Trust Hub on Apr 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches E2E frameworks and browser binaries (Playwright, Cypress, etc.) using package managers like npm, pip, poetry, and bundle from well-known registries.
  • [COMMAND_EXECUTION]: Executes shell commands to detect environment configurations, manage dependencies, start application services, and run the resulting test suites. This includes the use of 'node -e' for in-place modification of 'package.json' files.
  • [PROMPT_INJECTION]: The skill processes untrusted project data, specifically 'plan/spec.md' and git diff outputs, to drive its test authoring logic. This creates a surface for indirect prompt injection where malicious instructions in these files could influence agent behavior. Evidence: 1. Ingestion points: Reads from task-specific spec files and git diffs. 2. Boundary markers: Absent. 3. Capability inventory: Access to Bash, Write, and Edit tools across the entire project. 4. Sanitization: Absent; the ingested content is used directly to generate and execute test code.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 22, 2026, 11:26 AM
Security Audit — agent-trust-hub — e2e