ctf-web-recon
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill methodology requires the agent to analyze data from external, untrusted web sources which could contain instructions intended to manipulate the agent. 1. Ingestion points: Analysis of HTTP response headers, HTML comments, and JavaScript files retrieved via http_request. 2. Boundary markers: No specific delimiters or instructions to ignore embedded prompts are provided. 3. Capability inventory: Includes reconnaissance and command execution capabilities. 4. Sanitization: No instructions for sanitizing external content are included. \n- [DATA_EXFILTRATION]: The skill directs the agent to identify and access sensitive configuration and source files on a target system. Evidence: Targeted paths include environment files (.env), version control metadata (.git), and backup archives (.bak, .zip). \n- [COMMAND_EXECUTION]: The skill instructs the agent to execute the 'githacker' command-line tool to automate the recovery of Git repositories if detected (SKILL.md, Section 1.2).
Audit Metadata