everything-claude-code-harness-optimization

Fail

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to clone a complete repository from https://github.com/affaan-m/everything-claude-code.git and install various Node.js and Python packages, including ecc-universal, ecc-agentshield, torch, and torchvision.
  • [REMOTE_CODE_EXECUTION]: Following the clone, the instructions mandate the execution of an unverified shell installer (./install.sh) and NPM install scripts (npm run install), which can perform arbitrary actions on the host system.
  • [COMMAND_EXECUTION]: The setup process includes a JavaScript utility (scripts/harness-setup.js) that uses execSync to run shell commands based on environment variables like ECC_HARNESS. Additionally, the skill configures persistent shell scripts (hooks) such as SessionStart.sh and SessionStop.sh that run automatically without explicit user approval for each execution.
  • [DATA_EXFILTRATION]: The skill implements a session management and memory system that captures session summaries and metrics to local storage (.claude/memory/). Given the presence of automated shell hooks and network-capable installers, this system could be leveraged to exfiltrate session data or environment variables to external servers.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
May 16, 2026, 01:51 PM
Security Audit — agent-trust-hub — everything-claude-code-harness-optimization