Implement State Hub v0.2: dependency graph, next-steps suggestions, design boundary

S0 — Design boundary formalised across all integration surfaces:
- TOOLS.md restructured with Design Boundary section, Sanctioned Write Tools,
  and Bootstrap-Only Tools (create_workstream, create_task) with explicit note
- project_claude_md.template and railiance CLAUDE.md updated with boundary note
  and get_next_steps() in session start protocol
- Global ~/.claude/CLAUDE.md updated accordingly

S1 — Workstream dependency graph:
- WorkstreamDependency model (directed edge, CASCADE on delete, unique pair constraint)
- Alembic migration 0b547c153153; script.py.mako added (was missing)
- REST API: POST/GET /workstreams/{id}/dependencies/, DELETE …/{dep_id} (hard delete)
- StateSummary open_workstreams enriched with depends_on/blocks lists
- MCP tools: create_dependency(), list_dependencies()
- Dashboard workstreams page: Dependencies section with relationship cards
- Seeded: custodian-agent-runtime → llm-shared-library + phase-0-operational-baseline

S2 — Suggesting Next Steps (sanctioned write use case #2):
- GET /state/next_steps derives suggestions from recently resolved decisions
  (→ first open task in same workstream) and cleared dependencies
  (→ first todo task in now-unblocked workstream)
- StateSummary.next_steps included on every summary call
- MCP tool: get_next_steps()
- Dashboard: "What's next?" card grid above Registered Projects

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

mcp_server/server.py

@@ -62,6 +62,12 @@ def _patch(path: str, body: dict) -> Any:
         return r.json()
 
 
+def _delete(path: str) -> None:
+    with _client() as c:
+        r = c.delete(path)
+        r.raise_for_status()
+
+
 # ---------------------------------------------------------------------------
 # Resources
 # ---------------------------------------------------------------------------
@@ -398,6 +404,70 @@ def update_workstream_status(workstream_id: str, status: str) -> str:
     return json.dumps(ws, indent=2)
 
 
+# ---------------------------------------------------------------------------
+# Next-steps suggestion tool (S2.3) — sanctioned write use case #2
+# ---------------------------------------------------------------------------
+
+@mcp.tool()
+def get_next_steps() -> str:
+    """Surface contextual next-action suggestions derived from hub state.
+
+    Returns suggestions based on:
+    - Recently resolved decisions → first open task in the same workstream
+    - Workstreams whose every dependency is now completed → first todo task
+
+    Each suggestion includes domain, workstream, task, and a plain-language
+    message. The hub surfaces *what* and *where* — the domain owns *how*.
+
+    This is one of the two sanctioned write-side use cases of the State Hub
+    (the other is resolve_decision). Suggestions are derived, not persisted.
+    """
+    return json.dumps(_get("/state/next_steps"), indent=2)
+
+
+# ---------------------------------------------------------------------------
+# Dependency graph tools (S1.4)
+# ---------------------------------------------------------------------------
+
+@mcp.tool()
+def create_dependency(
+    from_workstream_id: str,
+    to_workstream_id: str,
+    description: str | None = None,
+) -> str:
+    """Record that one workstream depends on another.
+
+    Semantics: from_workstream cannot fully proceed until to_workstream reaches
+    a satisfactory state.
+
+    Args:
+        from_workstream_id: UUID of the workstream that has the dependency
+        to_workstream_id: UUID of the workstream it depends on
+        description: optional human-readable explanation of the dependency
+    """
+    dep = _post(f"/workstreams/{from_workstream_id}/dependencies", {
+        "to_workstream_id": to_workstream_id,
+        "description": description,
+    })
+    return json.dumps(dep, indent=2)
+
+
+@mcp.tool()
+def list_dependencies(workstream_id: str) -> str:
+    """Return all dependency edges touching a workstream (both directions).
+
+    The response distinguishes edges where this workstream is the dependent
+    (depends_on) from edges where it is the blocker (blocks).
+
+    Args:
+        workstream_id: UUID of the workstream to inspect
+    """
+    edges = _get(f"/workstreams/{workstream_id}/dependencies")
+    depends_on = [e for e in edges if e["from_workstream_id"] == workstream_id]
+    blocks = [e for e in edges if e["to_workstream_id"] == workstream_id]
+    return json.dumps({"depends_on": depends_on, "blocks": blocks}, indent=2)
+
+
 # ---------------------------------------------------------------------------
 # Entry point
 # ---------------------------------------------------------------------------