Skip to content

Security: aibackender/cognition-layer-framework

SECURITY.md

Security Policy

Supported Versions

Version Supported
1.0.x Yes

Reporting a Vulnerability

If you believe you have found a security vulnerability in the CLF specification or related repository materials, report it through GitHub private vulnerability reporting / GitHub Security Advisories for this repository.

  • Do not open a public issue for an unpatched vulnerability.
  • Include the affected version, impacted surface, reproduction details, and expected risk.
  • If the vulnerability is in a third-party CLF implementation, report it to that implementation's maintainers unless the issue appears to stem from the CLF specification itself.

Response Expectations

  • Reports will be acknowledged within 5 business days of receipt.
  • An initial severity and scope assessment will follow within 10 business days.
  • Fix timelines depend on severity and exploitability and will be communicated to the reporter after triage.
  • Coordinated disclosure is preferred until mitigations are available.
  • If you do not receive an acknowledgment within 5 business days, please follow up by opening a GitHub Security Advisory directly so the report is not lost.

Scope

Security-sensitive areas in this repository and specification include:

  • Trust boundaries between the orchestrator, adapters, subordinates, and external sources (Section 6.3 of the specification)
  • Prompt injection and indirect prompt injection handling, including instruction priority enforcement in injectPrompt() and adapter hooks
  • Secret handling via secretRef, which must resolve through the host secret manager and must not be serialized into configuration exports, checkpoints, event payloads, or audit logs
  • Adapter authorization and session or tenant scoping, including SecurityContext enforcement and multi-tenant partitioning by tenantId + sessionId
  • Event publication and auditability of sensitive actions, including configuration changes, pattern imports and promotions, deletion actions, and external invalidation events
  • Webhook-based invalidation sources, which must validate request authenticity before generating framework events

Specification Note

This repository contains a specification, not a single production implementation. Some vulnerabilities may arise from how a host runtime interprets or implements the spec rather than from the spec text itself. Reporters are encouraged to note whether the issue appears to be a spec ambiguity or an implementation gap, as this affects remediation path and disclosure coordination.

There aren't any published security advisories