registry / @sema-agent/core / 1.302.0

@sema-agent/core@1.302.0

Stateless, task-oriented AI agent core

AI Security Review

scanned 2d ago · by lpm-firewall-ai

No confirmed malicious install-time or import-time behavior exists. Runtime shell, workflow, and web capabilities are explicit agent-framework features rather than a hidden payload or persistence path.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

Consumer explicitly configures and invokes agent tools or workflow execution.

Impact

No unconsented exfiltration, persistence, destructive action, or AI-agent control-surface mutation was confirmed.

Mechanism

User-configured agent execution, web fetch, and workflow orchestration.

Rationale

The critical scanner signal is a false positive from static prompt text, while direct inspection shows no lifecycle payload, credential harvesting, exfiltration, or foreign agent-control mutation. The exposed risky primitives are package-aligned runtime features with visible guards and explicit invocation paths.

Evidence

package.jsondist/core/auto-mode-prompt-assets.jsdist/core/auto-promote.jsdist/engine/execution-env/node-execution-env.jsdist/tools/web.jsdist/orchestration/dev-vm-script-runner.jsdist/orchestration/run-workflow-tool.jsdist/server/http.js

Decision evidence

public snapshot

@sema-agent/core@1.302.0

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code