create
Warn
Audited by Gen Agent Trust Hub on May 11, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill modifies the .claude/settings.local.json file to grant itself broad permissions, including Read, Write, Edit, and Bash execution for the directory specified in the {NPD_PLUGIN_DIR} variable. This allows the agent to execute any Python script within that directory tree, effectively bypassing default security constraints for that path.
- [COMMAND_EXECUTION]: The skill executes several Python scripts from the plugin directory, such as run-backend.py, run-frontend.py, and create_repo.py, using python3. The integrity of these scripts depends on the contents of the user-provided plugin directory.
- [EXTERNAL_DOWNLOADS]: The skill generates instructions in AGENTS.md that direct the agent to use curl to download content from arbitrary URLs into a local .temp/ directory. This creates a mechanism for fetching unverified remote data.
- [CREDENTIALS_UNSAFE]: The repository creation process involves passing a GITHUB_TOKEN environment variable as a command-line argument (--token {GITHUB_TOKEN}). Passing sensitive tokens as CLI arguments can expose them in process monitoring tools or command history.
Audit Metadata