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/routing.toml

@@ -0,0 +1,47 @@
+# routing.toml — model routing by tier.
+# One place to change which model powers each tier. Read by aos-sync and the
+# Hermes orchestrator loop. Verify tool-calling quality per model before
+# trusting it with tools.
+
+[tiers.T1]
+description = "Orchestrator, architect, reviewer, security — strongest available Claude"
+model = "claude-sonnet-4-5"
+max_tokens_per_call = 200000
+allowed_tool_risk = ["low", "medium", "high"]
+
+[tiers.T2]
+description = "Department leads, complex workers"
+model = "claude-sonnet-4-5"
+max_tokens_per_call = 120000
+allowed_tool_risk = ["low", "medium"]
+
+[tiers.T3]
+description = "Bulk workers: classification, drafts, summaries, reports"
+model = "glm-4.5"
+fallback_model = "deepseek-chat"
+max_tokens_per_call = 60000
+allowed_tool_risk = ["low"]
+# Rule: no T3 model touches a risk:high tool without a T1/T2 reviewer pass.
+requires_reviewer_for_high_risk = true
+
+# Provider configuration — endpoints and which personal-data usage is allowed.
+# Personal data (client PII) may only be sent to providers marked
+# personal_data_allowed = true AND with an AVV (Art. 28 DSGVO) in place.
+
+[providers.claude]
+endpoint = "anthropic"
+personal_data_allowed = true
+avv_status = "ok"
+notes = "EU data residency via AWS eu-west-1 / eu-central-1."
+
+[providers.glm]
+endpoint = "zai"
+personal_data_allowed = false
+avv_status = "pending_legal_review"
+notes = "Use for non-PII bulk tasks only (code, public web content, summarization of anonymized data). Do NOT route client personal data here until AVV is signed."
+
+[providers.deepseek]
+endpoint = "deepseek"
+personal_data_allowed = false
+avv_status = "pending_legal_review"
+notes = "Same restriction as glm — non-PII bulk only."