Skip to content

Security: a5c-incubator/AgentPowerShell

SECURITY.md

Security Model

AgentPowerShell is designed to reduce risk around automated shell execution by enforcing policy, centralizing approval, and recording activity.

Core Controls

  • Policy-based command, file, environment, and network decisions
  • Session tracking and expiry
  • Authentication modes for local and service-facing use cases
  • Approval integration points for operator escalation
  • Event logging and reporting
  • LLM proxy redaction and usage tracking
  • MCP tool inventory, pinning, and cross-server detection
  • Workspace checkpoints for reversible changes

Current Scope

This repository provides a functional managed implementation and test coverage for the shared control plane. Platform-native enforcement components and signing/distribution flows should be treated as evolving work, not as production-complete hardening.

Reporting

Open a private security report through your normal project disclosure channel before filing a public issue.

There aren’t any published security advisories