feat: zeroclaw agentic agency architecture — agents, skills, Rust crates

- 18 agent specs across Software, Support, Marketing, Cross-cutting
  (orchestrator + 7 software + 5 support + 3 cross-cutting + leads-finder)
- 8 reusable skills (handoff, pr-review, dokploy-deploy, campaign-structure,
  content-calendar, ticket-triage, license-lifecycle, invoice-draft)
- 3 JSON schemas: handoff, deliverable, approval
- Policies: autonomy levels, tool risk classification, compliance (DSGVO/GoBD/UWG)
- routing.toml: model tiers T1/T2/T3 + provider config (PII gating)
- Rust workspace (5 crates, all compile + 16 tests pass):
  - aos-mcp: tool registry + risk tagging
  - aos-policy: autonomy levels, budget enforcement, approval requests
  - aos-license: ed25519 license issue/verify (real crypto + CLI binary)
  - aos-ledger: integer-cents money math, gapless numbering, VAT, audit chain
  - aos-sync: frontmatter validator (18/18 specs pass), CLI binary
- CLAUDE.md entry point with architecture overview + roadmap

agency-agents/schemas/handoff.json

@@ -0,0 +1,66 @@
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "$id": "https://zeroclaw.dev/schemas/handoff.json",
+  "title": "AgentHandoff",
+  "description": "Structured task delegation between agents. Every delegation uses this shape so messages stay debuggable. Free-form chat between agents is avoided.",
+  "type": "object",
+  "required": ["task_id", "from", "to", "goal", "definition_of_done"],
+  "additionalProperties": false,
+  "properties": {
+    "task_id": {
+      "type": "string",
+      "description": "AOS task ID (e.g. AOS-1423). Ties all tool calls, logs and approvals together.",
+      "pattern": "^AOS-\\d+$"
+    },
+    "from": {
+      "type": "string",
+      "description": "Agent name handing off (e.g. 'orchestrator', 'ads-lead').",
+      "minLength": 1
+    },
+    "to": {
+      "type": "string",
+      "description": "Agent name receiving the task.",
+      "minLength": 1
+    },
+    "goal": {
+      "type": "string",
+      "description": "One-sentence statement of what 'done' looks like, in business terms.",
+      "minLength": 1
+    },
+    "inputs": {
+      "type": "array",
+      "description": "References to inputs the receiving agent can read (AOS IDs, asset folders, brief IDs).",
+      "items": { "type": "string" }
+    },
+    "constraints": {
+      "type": "object",
+      "additionalProperties": false,
+      "properties": {
+        "budget_eur": { "type": "number", "minimum": 0, "description": "Hard spend cap in EUR for this task. 0 = no spend allowed." },
+        "max_tokens": { "type": "integer", "minimum": 0, "description": "Token cap for this task's agent calls." },
+        "deadline": { "type": "string", "format": "date-time" },
+        "autonomy": { "enum": ["L0", "L1", "L2"], "description": "Overrides the agent's default autonomy for this task. L3 not used initially." }
+      }
+    },
+    "definition_of_done": {
+      "type": "array",
+      "description": "Checklist of concrete, verifiable outcomes. Empty array is invalid.",
+      "minItems": 1,
+      "items": { "type": "string" }
+    },
+    "output_format": {
+      "enum": ["aos_deliverable", "report", "draft", "review_verdict", "data"],
+      "description": "How the result must be shaped so the caller can consume it without re-parsing."
+    },
+    "needs_approval_for": {
+      "type": "array",
+      "description": "Specific actions in this task that MUST pause for human approval (spend, publish, send, revoke, delete, deploy...).",
+      "items": { "type": "string" }
+    },
+    "context_refs": {
+      "type": "array",
+      "description": "Optional AOS records to load into context (client_id, project_id, contract_id).",
+      "items": { "type": "string" }
+    }
+  }
+}