Skip to content

v1.2.0 — Target-Scoped Reporting, Automated Native Setup & Dynamic Scope

Choose a tag to compare

@Zyrexnn Zyrexnn released this 20 Aug 09:44
· 103 commits to main since this release

🛡️ Cybermes v1.2.0

Target-Scoped Reporting, Automated Native Setup & Dynamic Scope Engine

Cybermes v1.2.0 introduces major enhancements to output organization, local host usability without Docker, anti-hallucination prompt stability, and dynamic operator authorization.


🚀 Key Highlights & New Features

1. 📁 Target-Scoped Reporting & Isolation

  • Organized Directory Hierarchy: All assessment deliverables are now isolated per target slug (reports/<target_slug>/), preventing cross-target overwrites and clutters.
    • reports/<target_slug>/findings/: Markdown reports per confirmed vulnerability.
    • reports/<target_slug>/pocs/: Standalone, reproducible Python/cURL PoC scripts.
    • reports/<target_slug>/evidence/: Raw HTTP request/response traces and screenshots.
    • reports/<target_slug>/SUMMARY.md: High-level executive finding matrix with severity counts.
    • reports/<target_slug>/metadata.json: Machine-readable scan metadata for dashboard integration.

2. 📊 Built-In Report Aggregator (tools/aggregate_reports.py)

  • Automated CLI utility to parse finding markdown files, extract CWE, CVSS v3.1, severity levels, and affected endpoints, and compile clean summary matrices.
  • Usage: python3 tools/aggregate_reports.py <TARGET_SLUG> or python3 tools/aggregate_reports.py --all.

3. ⚡ 1-Click Native Host Installer (setup.sh)

  • Full support for running natively on Linux/macOS without Docker.
  • Automated virtual environment setup (venv), dependency installation via requirements.txt, Playwright Chromium configuration, and Node.js MCP server registration.
  • Dynamic Path Resolution: env.sh, hermes, and bin/hermes now dynamically resolve $CYBERMES_DIR, eliminating hardcoded user paths.

4. 🧠 Anti-Hallucination & Anti-Refusal System Prompt

  • Optimized system prompt architecture in .hermes/config.yaml grounded strictly on tool output verification (stdout/stderr).
  • Eliminated ambiguous trigger words that caused false-positive model safety refusals.
  • Dynamic Scope Override: Full support for wildcard targets (targets: ["*"]) in scope.yaml — any target specified by the operator is immediately authorized without manual configuration file edits.

5. 🔑 Automated File Permissions & Docker Bind Mounts

  • Integrated umask 000 and default POSIX ACL configuration so all artifacts generated inside Docker are immediately readable and editable on the host editor without permission errors.
  • Added live bind mounts for skills/ and tools/ in docker-compose.yml for instant hot-reloading with docker compose restart.

👥 Contributors & Co-Authors