claude-code
Pass
Audited by Gen Agent Trust Hub on Aug 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONCREDENTIALS_UNSAFEREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The documentation provides instructions for installing the Claude Code CLI and associated SDKs from official Anthropic registries (npm and pip).
- [COMMAND_EXECUTION]: Includes numerous examples of shell commands used for feature implementation, debugging, and project management, as well as hook configuration.
- [CREDENTIALS_UNSAFE]: Provides guidance on managing ANTHROPIC_API_KEY and other environment variables, including security recommendations for .env files and secret rotation.
- [REMOTE_CODE_EXECUTION]: Explains how to extend the tool's functionality through plugins and Model Context Protocol (MCP) servers which involve script execution and remote tool integration.
- [DATA_EXFILTRATION]: Describes legitimate communication with official Anthropic API endpoints and external integrations like GitHub and Brave Search for search and repository access.
- [PROMPT_INJECTION]: Instructs the agent to assume the role of an expert by identifying topics and loading specific documentation references from the provided library.
- [INDIRECT_PROMPT_INJECTION]: Documents capabilities that ingest untrusted data such as codebase scans and CI logs. It identifies ingestion points in slash commands, provides an inventory of shell and filesystem capabilities, and highlights sanitization features like sandboxing and input validation in hooks.
Audit Metadata