GateKeeper6

GateKeeper6 is a static (non-executing) safety scanner for Python-focused repositories intended for use in MCEN/NIPR-like environments.

It helps produce a repeatable, readable audit by scanning source code for patterns that increase MCEN operational risk, especially:

• Non-loopback network communications (exfiltration/C2/unauthorized connections)
• Persistence / system modification behaviors (scheduled tasks, autoruns, system path writes)
• Dynamic code execution / loading (eval/exec, dynamic imports)
• Unsafe deserialization (pickle/marshal)
• Risky process execution (subprocess, shell=True, suspicious tooling)

The default profile is practical mode: it fails only on confirmed disallowed behavior and otherwise produces a CONDITIONAL_PASS with items requiring human review.

Getting started

• Scan a repo/dir: python3 -m mcen_scan /path/to/repo
• Scan current directory: python3 -m mcen_scan .
• Reports written to: ./mcen_audit/report.md and ./mcen_audit/report.json
• Manifest (per-file SHA-256): ./mcen_audit/manifest.json

Profiles

• Practical (default): --profile mcen_practical
  • Loopback (localhost, 127.0.0.1, ::1) is treated as allowed.
  • Unknown/dynamic network destinations are flagged for review (not auto-fail).
• Strict: --profile mcen_strict
  • Unknown/dynamic network destinations and non-allowlisted subprocess are treated more conservatively.

Subprocess allowlist (optional)

Use this to document and suppress benign subprocess review findings (does not suppress network/persistence/dynamic-exec findings):

• Example file: docs/allowlist.example.json
• Run with allowlist: python3 -m mcen_scan . --allowlist docs/allowlist.example.json

Network assurance (priority workflow)

GateKeeper6 supports two complementary ways to build strong evidence that a repo does not attempt non-loopback network access:

1. Static network evidence (default): detects common network APIs and call sites (NET* findings) without executing code.
2. Runtime egress harness (optional): runs a Python command under an egress-blocking harness that logs and blocks non-loopback connection attempts.

Runtime egress harness (optional)

This is supplemental, scenario-based evidence: it reflects the specific command/config/inputs you executed.

• Example (run repo entrypoint under harness):
  • python3 -m mcen_scan egress --output-dir ./mcen_audit . -- python3 your_entrypoint.py --arg value
• Output files:
  • mcen_audit/egress_log.jsonl (events)
  • mcen_audit/runtime_egress.json (summary)

See docs/RUNTIME_EGRESS_HARNESS.md for details and limitations.

What it scans

Default file types:

• Python: *.py (AST-based static analysis)
• PowerShell: *.ps1
• Batch: *.bat, *.cmd
• Shell: *.sh
• VBScript: *.vbs

It does not import or execute scanned code.

Docs

• Specification and rulepack: docs/MCEN_Python_Safety_Scanner_SPEC.md
• Milestones/implementation notes: docs/IMPLEMENTATION_PLAN.md
• Runtime harness usage: docs/RUNTIME_EGRESS_HARNESS.md

Development

• Tests: python3 -m unittest -q