claude-code-expert
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is purely instructional and documents the official Claude Code CLI by Anthropic. It provides guidance on configuration, workflows, and troubleshooting without executing any automated malicious actions.
- [SAFE]: Includes explicit security warnings regarding the use of the
--dangerously-skip-permissionsflag, advising that it should never be used in production or shared environments. - [SAFE]: Recommends security best practices such as denying high-risk commands (e.g.,
sudo,rm -rf, orcurlinstallers) within the.claude/settings.jsonpermissions configuration. - [SAFE]: Explicitly warns the agent not to print, record, or log API keys or sensitive credentials.
- [INDIRECT_PROMPT_INJECTION]: The skill documents the ingestion of project-specific context files (like
CLAUDE.mdand memory directories). It mitigates risk by: - Recommending explicit boundary markers and context-focus for sub-agents.
- Warning against executing code from untrusted PRs or external sources.
- Advising the use of isolated environments (sandboxes/CI) for automated tasks.
Audit Metadata