config-security-scan
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute commands such as find, grep, and cat. These are used to identify configuration files and search for sensitive patterns like API keys, tokens, and unsafe settings (e.g., dangerouslySkipPermissions) within the .claude/ directory. This activity is consistent with the skill's stated purpose of security auditing.
- [PROMPT_INJECTION]: The skill analyzes the content of local configuration files, which serves as a surface for indirect prompt injection. Malicious instructions embedded in these files could potentially influence the agent's behavior during the audit. 1. Ingestion points: Various files within the .claude/ directory (.json, .md, .yml, .js, .ts). 2. Boundary markers: The scan procedure lacks explicit delimiters or instructions to ignore natural language commands found within the scanned files. 3. Capability inventory: The skill utilizes Bash, Read, Grep, and Glob tools, which include shell execution capabilities. 4. Sanitization: The content read from the configuration files is processed without explicit sanitization or filtering.
Audit Metadata