feat: add production hardhat deployment scripts with safety checks

.agents/economy/ledger.json

@@ -0,0 +1,48 @@
+{
+  "version": "1.0.0",
+  "created_at": "2025-11-30T00:00:00Z",
+  "reset_daily": true,
+  "transactions": [],
+  "balances": {
+    "claude-architect": {
+      "input_tokens": 500000,
+      "output_tokens": 100000,
+      "usd_budget": 5.00,
+      "usd_spent": 0.00,
+      "tasks_completed": 0,
+      "subagents_spawned": 0
+    },
+    "codex-reviewer": {
+      "input_tokens": 200000,
+      "output_tokens": 50000,
+      "usd_budget": 2.00,
+      "usd_spent": 0.00,
+      "tasks_completed": 0,
+      "subagents_spawned": 0
+    },
+    "gemini-tester": {
+      "input_tokens": 200000,
+      "output_tokens": 50000,
+      "usd_budget": 1.50,
+      "usd_spent": 0.00,
+      "tasks_completed": 0,
+      "subagents_spawned": 0
+    },
+    "glm-auditor": {
+      "input_tokens": 200000,
+      "output_tokens": 50000,
+      "usd_budget": 0.50,
+      "usd_spent": 0.00,
+      "tasks_completed": 0,
+      "subagents_spawned": 0
+    }
+  },
+  "pricing": {
+    "claude-opus": { "input_per_1k": 0.015, "output_per_1k": 0.075 },
+    "claude-sonnet": { "input_per_1k": 0.003, "output_per_1k": 0.015 },
+    "claude-haiku": { "input_per_1k": 0.00025, "output_per_1k": 0.00125 },
+    "gpt-4": { "input_per_1k": 0.03, "output_per_1k": 0.06 },
+    "gemini-pro": { "input_per_1k": 0.00125, "output_per_1k": 0.00375 },
+    "glm-4": { "input_per_1k": 0.001, "output_per_1k": 0.002 }
+  }
+}

.agents/inbox/claude/ready.md

@@ -0,0 +1,5 @@
+ready
+
+agent: claude-architect
+timestamp: 2025-11-30T12:00:00Z
+status: active

.agents/manifest.json

@@ -0,0 +1,47 @@
+{
+  "version": "1.0.0",
+  "project": "xln",
+  "agents": {
+    "claude-architect": {
+      "votes": 2,
+      "role": "lead",
+      "can_spawn": true,
+      "description": "Lead developer, architecture decisions, can spawn subagents"
+    },
+    "codex-reviewer": {
+      "votes": 1,
+      "role": "reviewer",
+      "can_spawn": false,
+      "description": "Code review, security audit, bug finding"
+    },
+    "gemini-tester": {
+      "votes": 1,
+      "role": "tester",
+      "can_spawn": false,
+      "description": "Testing, edge cases, verification"
+    },
+    "glm-auditor": {
+      "votes": 1,
+      "role": "auditor",
+      "can_spawn": false,
+      "description": "Compliance, style, documentation"
+    }
+  },
+  "consensus": {
+    "threshold": 3,
+    "human_veto": true,
+    "timeout_minutes": 30
+  },
+  "polling": {
+    "interval_seconds": 60,
+    "max_retries": 3
+  },
+  "human_commands": {
+    "y": "approve & continue",
+    "n": "reject (explain why)",
+    "?": "show status",
+    "!": "emergency stop",
+    "1-9": "pick option",
+    " ": "skip/next"
+  }
+}

.agents/multiagent.md

@@ -0,0 +1,23 @@
+# NOT ACTIVE: Multi-Agent Protocol Extension Notes
+
+Status: inactive reference document.
+
+Do not use this as active workflow.
+Use `agents/workflow.md` as the canonical protocol.
+
+## Why Inactive
+- Too large and high-overhead for daily execution.
+- Includes optional economy and verbose papertrail policies.
+- Not fully tool-enforced.
+
+## Useful Ideas Preserved in Active Protocol
+- Human shortcuts: `y/n/?/!/1-9`
+- Optional queue/inbox mode
+- Atomic task claim pattern
+- Emergency stop behavior
+
+## Future Extension Candidates
+- Task queue schemas (`queue/task-*.md`)
+- Lightweight status snapshots per agent
+- Optional remote sync hooks
+- Optional voting records for major merges

.agents/papertrail/2025-12-01/council-1764586492011.md

@@ -0,0 +1,78 @@
+# Council Deliberation @ 2025-12-01T10:53:45.536Z
+
+## Query
+undefined
+
+## Stage 1: Individual Responses
+### qwen3-coder:latest
+Hello! I'm here to help you with any questions or topics you'd like to discuss. What would you like to explore or learn about today?
+
+### huihui_ai/qwen3-abliterated:235b
+It seems like you're looking for a step-by-step explanation or solution! Could you clarify the specific problem or topic you'd like help with? For example:  
+- A math or physics problem?  
+- An explanation of a concept?  
+- Something else entirely?  
+
+Let me know, and I’ll guide you through it! 😊
+
+### gpt-oss:120b
+It looks like your message came through empty. Could you let me know what you’d like help with? I'm here to answer questions, brainstorm ideas, troubleshoot problems, or anything else you need. Just let me know!
+
+## Stage 2: Peer Reviews
+### qwen3-coder:latest
+RANKINGS:
+- Response A: 8/10
+- Response C: 7/10
+
+BEST: Response A
+
+REASONING: Response A is the better of the two because it directly addresses the core issue - the empty message - with appropriate humility and openness. It acknowledges the problem clearly ("It looks like your message came through empty") while maintaining a helpful, non-confrontational tone. The response is concise, practical, and focuses on what the user actually needs: assistance with their actual question.
+
+Response C, while well-intentioned, makes an assumption that the user is seeking a "step-by-step explanation or solution" when the actual problem is that no message was received. This creates a mismatch between the response and the actual situation. Additionally, Response A is more universally applicable - it works whether the user needs help with a specific problem or just wants to communicate. Response A also demonstrates better self-awareness by recognizing it's dealing with an empty message rather than trying to guess what that message might contain.
+
+The main improvement for Response C would be to first acknowledge the empty message before asking for clarification, rather than immediately assuming what kind of help the user needs. Response A avoids this assumption entirely and simply asks for clarification, which is more appropriate given the circumstances.
+
+### gpt-oss:120b
+**RANKINGS**  
+- Response B: **6/10**  
+- Response C: **8/10**  
+
+**BEST:** Response C  
+
+**REASONING:**  
+The original query is simply the word “undefined,” which provides no concrete direction. The most appropriate immediate action is to request clarification so the user can be guided toward a useful answer.
+
+- **Accuracy:** Both replies correctly avoid giving misinformation and instead ask the user to specify their need. Response C’s acknowledgment that the user may be looking for a “step‑by‑step explanation or solution” is slightly more on point, earning it a higher accuracy rating.  
+
+- **Insight:** Response C demonstrates greater insight by anticipating likely categories of follow‑up (math/physics problem, concept explanation, etc.) and offering concrete examples. This helps the user quickly understand what kind of clarification is useful. Response B is polite but more generic.  
+
+- **Clarity:** Both messages are clear and easy to understand. However, Response C’s bullet‑point list makes the request for clarification even more explicit, which marginally improves its clarity score.  
+
+Overall, Response C is more helpful because it not only asks for clarification but also guides the user toward the type of information that would allow the assistant to provide a substantive answer.
+
+### huihui_ai/qwen3-abliterated:235b
+RANKINGS:
+- Response A: 9/10 (Accuracy: 10, Insight: 9, Clarity: 10)  
+- Response B: 8/10 (Accuracy: 9, Insight: 7, Clarity: 9)  
+
+BEST: Response A  
+
+REASONING:  
+1. **Accuracy**: Both responses correctly address the undefined query by prompting the user for clarification. Response A edges slightly higher by explicitly acknowledging the "undefined" issue ("your message came through empty") rather than just offering help.  
+2. **Insight**: Response A provides specific examples of assistance areas (e.g., brainstorming, troubleshooting), making it more actionable and informative. Response B is generic, offering no concrete direction beyond a broad "explore or learn."  
+3. **Clarity**: Both are clear, but Response A’s structured phrasing ("answer questions, brainstorm ideas, troubleshoot problems") and friendly tone ("I’m here to...") enhance readability. Response B is slightly less dynamic, relying on a standard prompt.  
+
+**Improvements**: Response B could adopt Response A’s specificity (e.g., listing common use cases) to better guide the user. Response A is already strong but might benefit from slightly more conciseness.
+
+## Stage 3: Chairman Synthesis (gpt-oss:120b)
+It looks like your message came through empty (or as the word “undefined”), so I’m not sure what you’d like help with. Could you let me know what you’re interested in or what problem you’d like to solve?
+
+Here are some common ways I can assist:
+
+- **Answering questions** – facts, definitions, how‑things‑work, etc.  
+- **Explaining concepts** – math, physics, programming, history, language, etc.  
+- **Brainstorming ideas** – project topics, writing prompts, design suggestions.  
+- **Troubleshooting** – debugging code, fixing a workflow, solving a technical issue.  
+- **Step‑by‑step guidance** – solving a problem, learning a new skill, planning a task.
+
+Just tell me which of these (or anything else) you’d like to explore, and I’ll dive right in!

.agents/profiles/claude-architect.md

@@ -0,0 +1,40 @@
+# agent: claude-architect
+
+## identity
+model: anthropic/claude-opus-4-5
+role: lead
+codename: claude-architect
+specialty: architecture, implementation, coordination
+
+## capabilities
+- full read/write access to /Users/zigota/xln
+- can run: bun, git, any bash command
+- can spawn: up to 2 concurrent subagents
+- can create: tasks, votes, reviews
+
+## responsibilities
+- lead development on xln
+- coordinate other agents
+- spawn subagents for verification
+- break ties in disputes
+
+## communication style
+- concise, technical
+- uses line numbers for code refs
+- prefers bullet points
+- respects token budgets
+
+## availability
+timezone: UTC
+active_hours: always (async)
+
+## budget
+see economy/ledger.json
+- 500k input / day
+- 100k output / day
+- $5.00 equiv / day
+
+## signed
+agent: claude-architect
+date: 2025-11-30
+status: active

.archive/2019_docs/.gitignore

@@ -0,0 +1 @@
+.DS_Store