Skip to content

Security: nessos666/coding-tentacle

Security

SECURITY.md

Security Policy — Coding Tentacle

Supported Versions

Version Supported
11.0.x ✅ Active support
< 11.0 ❌ End of life

Safety Guarantees

Coding Tentacle is a read-only safety layer. It NEVER:

  • Modifies files on disk
  • Executes code outside sandbox (/tmp/ct_sandbox_*)
  • Pushes, commits, or creates PRs
  • Auto-applies patches without human approval

Reporting a Vulnerability

Email: nessos6666@gmail.com

PGP: Not yet configured. Will be added.

Response time: 48 hours.

Security Architecture

CT employs 5 defense layers:

  1. TrojanSourceScanner — Detects Unicode Bidi, homoglyph, and zero-width attacks
  2. ASTSecurityAnalyzer — Structural Python code analysis (eval, exec, dangerous imports)
  3. SecurityStore — 10 CWE-mapped detection patterns (SQL injection, XSS, path traversal)
  4. InhibitoryControl — 5-level safety gate: GO / HOLD / BLOCK / ESCALATE / ASK_CONTEXT
  5. ApprovalGate — Human-in-the-loop final authorization

Absolute VETO: SafetyBrain BLOCK can NEVER be overridden by any other component.

Known Limitations

  • AST scan is Python-only (3.10+)
  • No runtime sandbox escape detection (relies on OS process isolation)
  • Droste code graph may be stale if not re-indexed after structural changes

Responsible Disclosure

We follow the Contributor Covenant. Please do NOT open public issues for security vulnerabilities.

There aren't any published security advisories