Skills
skills/aradotso/security-skills/vibe-pentest-ai-security-testing/Gen Agent Trust Hub

vibe-pentest-ai-security-testing

Fail

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs the agent to clone and run code from external, unverified repositories (github.com/ok-helloworld/vibe-pentest). This pattern facilitates the execution of arbitrary code from a remote source on the local system.
  • [COMMAND_EXECUTION]: The skill makes extensive use of the subprocess module to execute local Python scripts like fingerprint.py and run_katana.py. Since these scripts are downloaded from an external source, their behavior is not controlled by the skill definition.
  • [EXTERNAL_DOWNLOADS]: Installation requires fetching multiple Python packages and external binary tools (Katana crawler) from various third-party domains and registries.
  • [DATA_EXFILTRATION]: The skill contains logic to extract and save sensitive browser data, including session cookies and local storage tokens, using the Playwright automation library.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its processing of untrusted web content.
  • Ingestion points: Data from external web crawls is stored in workspace/crawl_summary.json and workspace/targets.txt and then processed by AI agents.
  • Boundary markers: No explicit delimiters or isolation instructions are provided for handling untrusted crawler data.
  • Capability inventory: The agent can execute scripts via subprocess.run and perform network requests using httpx.
  • Sanitization: No sanitization or validation of ingested web data is mentioned before it is processed by the AI agents.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jun 12, 2026, 08:51 PM
Security Audit — agent-trust-hub — vibe-pentest-ai-security-testing