Security Policy
Security guidelines and policies for Trajectory IR.
Trajectory IR acts as the durable semantic and execution layer for autonomous AI agents. A compromise in this layer could lead to unauthorized tool execution, state manipulation, or the leakage of sensitive data (like PII or secrets).
This policy is tightly integrated with our Infrastructure Design, Contributing Guidelines, and Code of Conduct.
1. Supported Versions
Trajectory IR is currently in its Phase 1A / v0.1.x development cycle.
| Version | Supported | Notes |
|---|---|---|
| v0.1.x | :white_check_mark: | Active development (DBOS embedded backend) |
| < v0.1 | :x: | Historical (CAMI/CLOOP prototypes) |
2. Reporting a Vulnerability
Please do not report security vulnerabilities through public GitHub issues.
If you discover a security vulnerability, please send an e-mail to the core maintainers: siddharthagithub0007@gmail.com (or project owner email).
We will acknowledge receipt of your vulnerability report within 48 hours. Please adhere to the Code of Conduct during this process—public zero-day drops or harassment of maintainers over patches are strict violations of our community standards.
3. Scope of Security Concerns (Architecture Specific)
Based on the Infrastructure Blueprint and Master Spec, we are actively monitoring for vulnerabilities in the following planes:
A. Execution & Tool Safety Plane
- Safety Boundary Bypasses: Exploits that trick the system into classifying a
NON_IDEMPOTENT_WRITEtool asPUREorREAD_ONLY, bypassing the Fail-Closed default. - Block-and-Gate Evasion: Flaws that allow an interrupted non-idempotent tool to automatically retry without explicit human/policy resolution.
- Backend Injection: Any flaw in
drivers/durable-backend/dbos/that allows arbitrary code execution outside of the locked DBOS/Restate step wrapper context.
B. State & Durability Plane
- Seal Tampering: Vulnerabilities allowing a node payload to be mutated without breaking the RFC 8785 (JCS) + SHA256 identity hashing.
- Cache Poisoning (
k8s-fluidprofile): Exploits where a stale or poisoned Fluid Dataset FUSE mount can trick the runtime into bypassing the direct S3/MinIO CAS hash-verification fallback.
C. Data & Export Plane (.tir Packages)
- Sensitive Data Leakage: Flaws where nodes marked with the
SENSITIVEeffect class fail to be stripped or hashed properly during aredacted.tirpackage export.
4. Security Accountability for Contributors
As defined in our Contributing Guidelines:
- AI Generation Liability: If you use AI coding assistants (Antigravity IDE, Claude Code, ECC) to draft PRs, you, the human contributor, are 100% accountable for any security flaws they introduce. AI agents have zero built-in trust regarding security boundaries.
- Mandatory Security Reviews: Any pull request that modifies files in
pkg/effects/(tool safety mapping) orpkg/resume/(block-and-gate semantics) is automatically flagged for maximum scrutiny and requires sign-off from the Security-Review Agent and a human core maintainer.
Thank you for helping keep Trajectory IR safe and verifiable!