AgentGuard
Security scanner that vets AI-agent repos, MCP tool configurations, and prompt parameters before execution — enforcing 70+ SARIF-ready static analysis rules.
Source on GitHubThe Problem
Autonomous AI agents and MCP servers execute code, edit local files, and invoke external APIs. Without preflight policy enforcement, rogue tool calls or malicious repo instructions can compromise developer workstations.
The Solution
AgentGuard enforces strict security boundaries before agent execution:
- 70+ Static Rules: Audits tool signatures, prompt injections, and shell invocation boundaries.
- SARIF Integration: Generates standard SARIF reports compatible with GitHub Security tab and CI pipelines.
- Aegis Engine: Utilizes deterministic preflight policies to safeguard environment secrets and filesystem paths.
Next — 14