crawl
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes external reconnaissance binaries (
hakrawlerandgospider) through a Python wrapper. The script uses list-based arguments forsubprocess.run, which is a safe practice that prevents shell command injection. - [EXTERNAL_DOWNLOADS]: The installation script (
scripts/install_tools.sh) downloads and installs tools from third-party GitHub repositories (github.com/hakluke/hakrawlerandgithub.com/jaeles-project/gospider) using the Go package manager. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to crawl external websites and present the findings (URLs, subdomains, JS content) to the agent for analysis. This creates an attack surface where a malicious website could host prompt injection payloads intended to influence the agent's behavior when it reads the generated reports or logs.
- Ingestion points: External URL content retrieved by
hakrawlerandgospiderand stored inresults/andlogs/directories. - Boundary markers: None; external content is written directly to text, JSON, and Markdown files without specific delimiters to warn the agent about embedded instructions.
- Capability inventory: The skill environment allows file system writes, network requests (via tools), and execution of integrated scripts.
- Sanitization: No sanitization or filtering of the discovered URL strings or tool output is performed before presentation to the agent.
Audit Metadata