Skip to content

Commit 878ed92

Browse files
Ascendralclaude
andcommitted
Add SECURITY.md with disclosure policy and security architecture
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1 parent 6896ea4 commit 878ed92

1 file changed

Lines changed: 52 additions & 0 deletions

File tree

SECURITY.md

Lines changed: 52 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,52 @@
1+
# Security Policy
2+
3+
## Supported Versions
4+
5+
| Version | Supported |
6+
|---------|-----------|
7+
| 4.0.x | ✅ Current |
8+
| 3.0.x | ⚠️ Security fixes only |
9+
| < 3.0 | ❌ End of life |
10+
11+
## Reporting a Vulnerability
12+
13+
**Do not open a public GitHub issue for security vulnerabilities.**
14+
15+
Email **alex@zanderpinkdesign.com** with:
16+
17+
1. Description of the vulnerability
18+
2. Steps to reproduce
19+
3. Affected versions
20+
4. Impact assessment (if known)
21+
22+
You will receive an acknowledgment within **48 hours** and a detailed response within **5 business days**.
23+
24+
## Disclosure Policy
25+
26+
- We follow [coordinated disclosure](https://en.wikipedia.org/wiki/Coordinated_vulnerability_disclosure).
27+
- After a fix is released, we will publish a GitHub Security Advisory.
28+
- Credit will be given to reporters unless they request otherwise.
29+
30+
## Security Architecture
31+
32+
CORD Engine is itself a security product — it evaluates AI agent actions for safety. Its own security posture:
33+
34+
| Layer | Mechanism | Description |
35+
|-------|-----------|-------------|
36+
| **14-Dimension Risk Scoring** | Constitutional evaluation | Every proposed action scored across 14 risk dimensions (0–99) |
37+
| **Hard Blocks** | Non-overridable | Critical safety violations cannot be bypassed by any user or system |
38+
| **VIGIL Threat Scanner** | Proactive defense | Scans for prompt injection, canary tokens, trajectory anomalies |
39+
| **Red Team Validated** | 40 attack scenarios | 100% block rate across adversarial test suite |
40+
| **Zero Dependencies** | Self-contained | No npm dependencies — entire engine is standalone |
41+
42+
## Dependency Policy
43+
44+
- **Zero runtime dependencies** — CORD Engine has no npm dependencies
45+
- All evaluation logic is implemented from scratch
46+
- No network calls during evaluation (fully offline-capable)
47+
48+
## Supply Chain
49+
50+
- Published to npm with provenance
51+
- 482 tests across 13 test suites
52+
- Red team validation suite included in repository

0 commit comments

Comments
 (0)