ghost-scan-deps
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill downloads and executes an installation script from the author's official GitHub repository (https://raw.githubusercontent.com/ghostsecurity/wraith/main/scripts/install.sh) to set up the wraith scanner binary.
- [COMMAND_EXECUTION]: The skill uses shell commands to execute the wraith binary for dependency scanning and uses git commands to determine repository metadata for reporting.
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted third-party data to determine if vulnerabilities are exploitable, which creates a potential surface for indirect prompt injection.
- Ingestion points: Dependency lockfiles (e.g., package-lock.json, go.mod), scanner JSON output, and repository source code files accessed via Grep and Read tools in agents/analyze/analyzer.md.
- Boundary markers: The analyzer.md prompt instructs the agent to focus on specific exploitability criteria, but it does not employ strict delimiters or 'ignore' instructions for the source code it processes.
- Capability inventory: The skill utilizes the Task tool to orchestrate subagents, and subagents have access to Bash, Read, Grep, and Glob tools to perform their analysis.
- Sanitization: No pre-processing or sanitization is applied to the source code snippets or vulnerability data before they are evaluated by the AI agent.
Audit Metadata