collaborating-with-claude

Pass

Audited by Gen Agent Trust Hub on May 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The scripts/claude_bridge.py script executes the claude CLI via subprocess.Popen. It implements security best practices by using shell=False and passing arguments as a list, preventing shell injection at the bridge level.\n- [PROMPT_INJECTION]: The skill exposes a prompt injection surface as it passes user-provided text directly to the Claude Code CLI, which has capabilities including shell execution and file modification.\n
  • Ingestion points: Instructions are accepted through the --PROMPT parameter in scripts/claude_bridge.py.\n
  • Boundary markers: The script and its associated prompt templates do not employ explicit boundary markers or delimiters for user-provided data.\n
  • Capability inventory: The underlying claude tool can execute arbitrary shell commands, read/write files, and interact with the environment through MCP tools.\n
  • Sanitization: No input sanitization or validation is performed on the prompt text before transmission to the CLI.
Audit Metadata
Risk Level
SAFE
Analyzed
May 15, 2026, 07:30 AM
Security Audit — agent-trust-hub — collaborating-with-claude