@copilotkit/runtime@1.64.0
<img src="https://github.com/user-attachments/assets/0a6b64d9-e193-4940-a3f6-60334ac34084" alt="banner" style="border-radius: 12px; border: 2px solid #d6d4fa;" />
AI Security Review
scanned 3mo ago · by lpm-firewall-aiThe flagged primitives are runtime features of an AI integration package, activated by consumer configuration or requests. Source inspection found no lifecycle execution, covert payload execution, filesystem mutation, credential harvesting, or foreign AI-agent control-surface changes.
Decision evidence
public snapshot- Runtime telemetry captures request/instance metadata and a supplied public API key.
- Managed Intelligence client sends configured API calls with its configured bearer key.
- package.json has no preinstall/install/postinstall/prepare lifecycle hook.
- dist/index.cjs only loads package runtime modules on import; no shell, eval, or filesystem mutation found.
- dist/lib/telemetry-disclosure.cjs discloses telemetry and supports COPILOTKIT_TELEMETRY_DISABLED/DO_NOT_TRACK opt-out.
- dist/v2/runtime/handlers/intelligence/annotate.cjs handles explicit POST requests and forwards configured annotation data.
- dist/v2/runtime/intelligence-platform/client.cjs uses documented CopilotKit Intelligence endpoints configured by the caller.
Source & flagged code
2 flaggedPackage source references dynamic require/import behavior.
dist/langgraph.cjsView on unpkg · L1Package metadata claims a different repository identity while copied source loads a runtime dependency bridge.
dist/v2/runtime/handlers/intelligence/annotate.cjsView on unpkg · L6Findings
Affected versions and remediation
This report applies to @copilotkit/runtime@1.64.0.
See all versions of @copilotkit/runtime for other recorded verdicts.
- Review the evidence and your use of @copilotkit/runtime@1.64.0 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.