Repository navigation
Home
divyanshujethi edited this page Oct 13, 2026
·
5 revisions
Welcome to the official AgentShield Knowledge Base and Technical Architecture Wiki.
AgentShield is an enterprise-grade, local-first runtime security layer and AST shell sandbox designed specifically for autonomous AI coding agents (Claude Code, Cursor, Windsurf, OpenAI Codex, and Google Antigravity).
- Architecture-Overview: High-performance Aho-Corasick state automata, Shannon entropy heuristic scanner, and zero-collision SHA-256 surrogate token replacement.
-
Agent-Sandboxing-and-AST-Normalization: How AgentShield normalizes obfuscated shell commands and protects
.envand cloud credentials from entering LLM context. - CI-CD-Security-Gate-and-SARIF: Auditing pull requests in GitHub Actions with SARIF 2.1.0 reporting.
- MCP-Protocol-Proxy: Inline JSON-RPC interceptor for Model Context Protocol servers.
- Enterprise-Compliance-Attestation: Compliance mapping for SOC2 CC6.1, PCI-DSS 3.4, HIPAA §164.312, and ISO27001 A.8.28.
- Roadmap-and-Project-Board: Strategic milestones, backlog items, and project board organization.
- Official Website: https://agentshield.ritualdev.in
- Ecosystem Hub: DevShelf Directory
- Repository: RitualDev-Lab/AgentShield
-
npm / GitHub Packages:
@ritualdev_lab/agentshield
AgentShield is maintained by RitualDev Lab • Licensed under Apache-2.0 • Zero Telemetry Guaranteed.