Skills
Navigate
skills/reason-machines/security-skills/pentest-ai-agents/Gen Agent Trust Hub

pentest-ai-agents

Fail

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFEPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The installation instructions recommend executing a remote shell script directly from an untrusted source via a pipe to bash.
  • Evidence: curl -fsSL https://raw.githubusercontent.com/0xSteph/pentest-ai-agents/main/install.sh | bash in SKILL.md.
  • [REMOTE_CODE_EXECUTION]: The skill facilitates the installation of local model infrastructure by fetching a script from Ollama's official service.
  • Evidence: curl -fsSL https://ollama.com/install.sh | sh in SKILL.md.
  • [COMMAND_EXECUTION]: The skill documentation includes numerous examples of the agent executing powerful shell commands and third-party security tools, creating an extensive attack surface.
  • Evidence: Mentions of nmap, msfvenom, binwalk, evilginx2, and custom bash scripts (doctor.sh, findings.sh, handoff.sh) throughout SKILL.md.
  • [CREDENTIALS_UNSAFE]: The documentation provides examples for setting sensitive environment variables, which could lead to accidental exposure of production keys.
  • Evidence: export ANTHROPIC_API_KEY="sk-ant-..." in SKILL.md.
  • [PRIVILEGE_ESCALATION]: The skill documents and enables high-privilege operations, including the use of sudo and techniques for escaping container environments to the host system.
  • Evidence: sudo apt install nmap and docker -H unix:///var/run/docker.sock run -v /:/host -it alpine chroot /host /bin/bash in SKILL.md.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze a wide variety of untrusted external data formats, which could be used to influence agent behavior.
  • Ingestion points: The skill processes Nmap XML, Nuclei JSON, BloodHound data, firmware binaries, and PCAP files (SKILL.md).
  • Boundary markers: No explicit boundary markers or "ignore embedded instructions" directives are defined for these data sources.
  • Capability inventory: The skill has broad execution capabilities including network operations, file system writes, and shell command execution across all Tier 2 agents.
  • Sanitization: There is no evidence of sanitization or strict schema validation for these ingested files.
Recommendations
  • HIGH: Downloads and executes remote code from: https://ollama.com/install.sh, https://raw.githubusercontent.com/0xSteph/pentest-ai-agents/main/install.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 15, 2026, 08:57 PM
Security Audit — agent-trust-hub — pentest-ai-agents