crack-hashcat

Warn

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill provides instructions to access highly sensitive credential storage files on the local system, such as /etc/shadow, and Active Directory databases like ntds.dit.
  • [COMMAND_EXECUTION]: Multiple commands utilize sudo to acquire the elevated privileges necessary for reading restricted system files and directories.
  • [REMOTE_CODE_EXECUTION]: The provided CI/CD template in assets/ci-config-template.yml includes a command pattern that downloads an installation script from the tfsec repository on GitHub and executes it directly via the shell.
  • [EXTERNAL_DOWNLOADS]: The CI/CD configuration template describes the installation of several external security tools (Semgrep, Safety, Checkov, Bandit, Gitleaks, and Trivy) from public package registries and GitHub.
  • [PROMPT_INJECTION]: The skill instructions include processing of external data files, such as hash lists and JSON scan results, without explicit boundary markers or sanitization, exposing a surface for indirect prompt injection.
  • Ingestion points: SKILL.md (hash files, wordlists), assets/ci-config-template.yml (security scan results).
  • Boundary markers: Absent for all external file ingestion.
  • Capability inventory: Execution of hashcat, secretsdump.py, and shell commands across multiple files.
  • Sanitization: No evidence of validation or sanitization of input data before processing or interpolation.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 18, 2026, 11:00 AM
Security Audit — agent-trust-hub — crack-hashcat