claude-code-coding-agent
Pass
Audited by Gen Agent Trust Hub on Apr 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on executing the
claudecommand-line interface to perform tasks. It instructs the agent to use various flags such as--printfor headless execution and--permission-modeto manage tool permissions. - [COMMAND_EXECUTION]: The skill is designed to support user-provided wrappers or aliases for the
claudecommand, which involves executing arbitrary shell commands based on the local environment configuration. - [DATA_EXFILTRATION]: One example suggests redirecting process output to
/tmp/claude-code-agent.log. On multi-user systems, files in the/tmpdirectory are typically world-readable, which could lead to the unintended exposure of source code, logs, or session data to other users on the host.
Audit Metadata