claude-code-expert
Fail
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill references a hardcoded script path 'C:/Users/renat/skills/cred-omega/scripts/secret_scanner.py' within a configuration hook example. This path points to a specific user directory unrelated to the vendor, posing a risk of executing unverified local code from an unknown source.\n- [COMMAND_EXECUTION]: The skill instructs the agent and user to execute various external scripts (e.g., 'agent-orchestrator/scripts/scan_registry.py', 'context_manager.py', 'pre_task_check.py') that are not provided within the skill's own file set, creating a critical dependency on an unverified external environment.\n- [PROMPT_INJECTION]: The 'CLAUDE.md' section contains steering instructions (e.g., 'Protocolo Pre-Tarefa: Sempre rodar orchestrator antes de responder') that function as prompt injections to override the agent's default operational behavior.\n- [REMOTE_CODE_EXECUTION]: The skill includes a platform-level 'security-allowlist' for 'curl-pipe-bash', indicating a design intended to facilitate the execution of remote scripts directly into a shell session.\n- [COMMAND_EXECUTION]: The documentation encourages the use of the '--dangerously-skip-permissions' flag for the Claude Code CLI, which explicitly bypasses security confirmation prompts for sensitive operations like file system writes and command execution.\n- [EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the '@anthropic-ai/claude-code' package via npm. While this is a well-known service from a trusted organization (Anthropic), it establishes a pattern of installing external executable code.
Recommendations
- AI detected serious security threats
Audit Metadata