looplia-e2e

Warn

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes multiple shell scripts (e2e.sh, e2e-setup.sh, e2e-auto-discovery.sh) that perform build operations (bun run build), dependency management (bun x tsup), and direct execution of local CLI binaries.
  • [CREDENTIALS_UNSAFE]: The skill interacts with .env files containing sensitive API keys (ZENMUX_API_KEY). Specifically, the e2e-setup.sh script sources the .env file, and SKILL.md provides a troubleshooting command (cat .env | grep ZENMUX_API_KEY) that prints sensitive credentials to the console, exposing them to the agent context.
  • [PROMPT_INJECTION]: The skill contains explicit instructions and code to bypass platform-level safety mechanisms. It uses env -u CLAUDECODE and unset CLAUDECODE in the setup_test_env function to circumvent restrictions designed to prevent recursive or nested agent sessions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 18, 2026, 09:16 PM
Security Audit — agent-trust-hub — looplia-e2e