From 1061e9502d2f7863f90e32356e266ca7e913f59d Mon Sep 17 00:00:00 2001
From: GAP Promoter <bot@gitagent.sh>
Date: Wed, 27 May 2026 21:42:13 +0000
Subject: [PATCH] Add rjmurillo/ai-agents to the registry

---
 agents/rjmurillo__ai-agents/README.md     | 60 +++++++++++++++++++++++
 agents/rjmurillo__ai-agents/metadata.json | 14 ++++++
 2 files changed, 74 insertions(+)
 create mode 100644 agents/rjmurillo__ai-agents/README.md
 create mode 100644 agents/rjmurillo__ai-agents/metadata.json

diff --git a/agents/rjmurillo__ai-agents/README.md b/agents/rjmurillo__ai-agents/README.md
new file mode 100644
index 0000000..74692e6
--- /dev/null
+++ b/agents/rjmurillo__ai-agents/README.md
@@ -0,0 +1,60 @@
+# AI Agent System — rjmurillo/ai-agents
+
+A production-grade, coordinated multi-agent system for software development teams
+that want AI velocity **with** real governance.
+
+## What it does
+
+Ships **23+ specialized AI agents** for Claude Code (and 24 for GitHub Copilot CLI)
+that cover every phase of the software development lifecycle:
+
+| Role | Agent | Responsibility |
+|------|-------|----------------|
+| Coordination | `orchestrator` | Central hub; delegates to specialists, keeps context clean |
+| Discovery | `analyst` | Research, spikes, requirements, and feasibility |
+| Design | `architect` | System design, ADR authoring, tech-debt triage |
+| Build | `implementer` | Code generation guided by specs and ADRs |
+| Quality | `qa` | Testing strategy, Pester/pytest authoring, coverage |
+| Safety | `security` | Vulnerability detection, CWE/OWASP scanning |
+| Operations | `devops` | CI/CD pipelines, deployment, and observability |
+| Adversarial | `critic` | Pokes holes in other agents' outputs |
+| Governance | `session-init` / `session-end` | Session gates, memory persistence, handoffs |
+| Improvement | `reflect` | Weekly retrospectives and self-improvement tracking |
+
+## Key capabilities
+
+- **ADR-steered governance** — architecture decisions live in `.agents/architecture/ADR-*.md`; agents never break them without a new ADR proposal
+- **Explicit handoff protocols** — each agent passes a structured HANDOFF.md to the next, keeping the orchestrator context lean
+- **Cross-session memory** — via Serena MCP + Forgetful; citation-verified graph traversal with health reporting
+- **Review gates** — five validation checkpoints before a PR lands (spec, pre-PR script, security scan, QA, and ADR compliance)
+- **Skill-first routing** — any task that matches an installed skill invokes that skill's tested workflow first
+
+## Installation
+
+**Claude Code:**
+```
+/install-plugin rjmurillo/ai-agents
+```
+
+**GitHub Copilot CLI:**
+```
+/plugin marketplace add rjmurillo/ai-agents
+/plugin install project-toolkit@ai-agents
+```
+
+## Example usage
+
+```
+# Claude Code — run the orchestrator on a new feature
+Task(subagent_type="orchestrator", prompt="Build a rate-limiting middleware for our FastAPI service per ADR-017")
+
+# Trigger a security scan
+Task(subagent_type="security", prompt="Scan PR #234 for CWE-89 and OWASP A03")
+
+# Weekly retro
+Task(subagent_type="reflect", prompt="Run the weekly retrospective")
+```
+
+## License
+
+MIT — [github.com/rjmurillo/ai-agents](https://github.com/rjmurillo/ai-agents)
diff --git a/agents/rjmurillo__ai-agents/metadata.json b/agents/rjmurillo__ai-agents/metadata.json
new file mode 100644
index 0000000..d67fb18
--- /dev/null
+++ b/agents/rjmurillo__ai-agents/metadata.json
@@ -0,0 +1,14 @@
+{
+  "name": "ai-agents",
+  "author": "rjmurillo",
+  "description": "Coordinated multi-agent system for software development: 23+ specialized agents (orchestrator, architect, QA, security, devops) with ADR governance, session protocols, and cross-session memory.",
+  "repository": "https://github.com/rjmurillo/ai-agents",
+  "version": "1.0.0",
+  "category": "developer-tools",
+  "tags": ["multi-agent", "claude-code", "orchestrator", "software-development", "governance", "adr", "devops", "code-review", "mcp", "session-protocol"],
+  "license": "MIT",
+  "model": "claude-sonnet-4-6",
+  "adapters": ["claude-code", "system-prompt"],
+  "icon": false,
+  "banner": false
+}