Hi @ec-jrc maintainers,
Quick proposal: open detection-rule layer integration for screening malicious MCP servers passing through Aloha as the orchestrator.
Context: I maintain Agent Threat Rules (ATR, https://github.com/Agent-Threat-Rule/agent-threat-rules), an MIT-licensed detection-rule corpus for AI agent threats. 348 rules at v2.1.4 covering prompt injection, tool poisoning, context exfiltration, agent loop exploits. In production at Microsoft Agent Governance Toolkit, Cisco AI Defense (314-rule pack), MISP / CIRCL Luxembourg (merged 2026-05-10 by adulau), and OWASP Agent-Security-Regression-Harness. Operational cadence: 2 hours 16 minutes from MSRC Semantic Kernel CVE-2026-26030 disclosure (2026-05-07) to npm-published detection rules (2026-05-11).
Why Aloha specifically: as an MCP + A2A protocol orchestrator, Aloha sits exactly where malicious MCP servers would first surface to a downstream agent stack. The same threat shape that motivated Microsoft Copilot SWE Agent to open regression-test fixtures in microsoft/agent-governance-toolkit#1981 (2026-05-11) applies here, except Aloha is at the EU side of that pipeline.
What I am asking before any PR:
- Is Aloha actively accepting external contributions, or is it currently an internal-only JRC research codebase?
- If contributions are welcome, would a detection-rule loader extension (load ATR YAML rules at orchestrator startup, run them against MCP tool descriptions and tool-call payloads and LLM responses passing through Aloha) be in scope, OR is detection considered out of scope for the orchestrator layer?
- If different in scope but the orchestrator could emit structured threat events for downstream consumers, that pattern works too. We publish an Aloha-shape adapter; Aloha stays clean.
Honest about what I will NOT propose:
- No new license restrictions. ATR is MIT and works fine alongside Aloha's EUPL-1.2 in adapter pattern.
- No vendor binding. The integration is multi-vendor by design. ATR is one of several open corpora; garak, HarmBench, OWASP A-S-R-H all map similarly.
- No request that JRC endorse ATR. The goal is to make detection rules pluggable, with ATR as one possible loader and JRC retaining authority to add or reject loaders.
For context, this proposal is part of a Wave-1 outreach across CCCS Canada and ACSC Australia happening the same week. No urgency, just transparency about the broader plan.
Happy to draft a small concrete PR once the scope question is answered.
Repo: https://github.com/Agent-Threat-Rule/agent-threat-rules
Maintainer: Adam Lin, adam@agentthreatrule.org
Foundation: Panguard AI Inc. (Delaware C-Corp, filed 2026-05-12)
Hi @ec-jrc maintainers,
Quick proposal: open detection-rule layer integration for screening malicious MCP servers passing through Aloha as the orchestrator.
Context: I maintain Agent Threat Rules (ATR, https://github.com/Agent-Threat-Rule/agent-threat-rules), an MIT-licensed detection-rule corpus for AI agent threats. 348 rules at v2.1.4 covering prompt injection, tool poisoning, context exfiltration, agent loop exploits. In production at Microsoft Agent Governance Toolkit, Cisco AI Defense (314-rule pack), MISP / CIRCL Luxembourg (merged 2026-05-10 by adulau), and OWASP Agent-Security-Regression-Harness. Operational cadence: 2 hours 16 minutes from MSRC Semantic Kernel CVE-2026-26030 disclosure (2026-05-07) to npm-published detection rules (2026-05-11).
Why Aloha specifically: as an MCP + A2A protocol orchestrator, Aloha sits exactly where malicious MCP servers would first surface to a downstream agent stack. The same threat shape that motivated Microsoft Copilot SWE Agent to open regression-test fixtures in microsoft/agent-governance-toolkit#1981 (2026-05-11) applies here, except Aloha is at the EU side of that pipeline.
What I am asking before any PR:
Honest about what I will NOT propose:
For context, this proposal is part of a Wave-1 outreach across CCCS Canada and ACSC Australia happening the same week. No urgency, just transparency about the broader plan.
Happy to draft a small concrete PR once the scope question is answered.
Repo: https://github.com/Agent-Threat-Rule/agent-threat-rules
Maintainer: Adam Lin, adam@agentthreatrule.org
Foundation: Panguard AI Inc. (Delaware C-Corp, filed 2026-05-12)