feat(mcp): add get_domain_summary() for low-token domain session orientation

get_state_summary() returns ~10k tokens — too expensive for routine domain
repo sessions that only need their own workstreams and decisions.

New get_domain_summary(domain_slug):
- 5 targeted API calls: topics (filter), workstreams (topic+status), decisions
  (topic+pending), progress (topic, limit 5), repos (domain, slug+SBOM only)
- Returns: topic, active workstreams, blocking decisions, 5 recent events,
  repo SBOM status — all scoped to one domain
- Estimated ~80-90% token reduction vs get_state_summary()

get_state_summary() preserved unchanged for cross-domain / custodian sessions.
Updated its docstring to note the large response and point to get_domain_summary.

Template updated: Step 1 now calls get_domain_summary("{DOMAIN}") instead of
get_state_summary() + get_next_steps(). TOOLS.md updated with usage guidance.

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

mcp_server/server.py

@@ -24,6 +24,8 @@ mcp = FastMCP(
     instructions=(
         "Custodian State Hub: tracks topics, workstreams, tasks, decisions, and progress events. "
         "Start every session with get_state_summary() for orientation. "
+        "When working inside a single registered domain repo, prefer get_domain_summary(domain_slug) "
+        "— it returns the same actionable data scoped to that domain at ~10% of the token cost. "
         "All writes emit a progress_event automatically."
     ),
 )
@@ -120,10 +122,51 @@ def get_state_summary() -> str:
 
     Returns a full snapshot: topic/workstream/task/decision totals, blocking
     decisions, blocked tasks, open workstreams, and the 20 most recent events.
+
+    NOTE: This response is large (~10k tokens). When working inside a single
+    registered domain repo, use get_domain_summary(domain_slug) instead —
+    same actionable data scoped to one domain at ~10% of the token cost.
     """
     return json.dumps(_get("/state/summary"), indent=2)
 
 
+@mcp.tool()
+def get_domain_summary(domain_slug: str) -> str:
+    """Lightweight session orientation for a single domain.
+
+    Use this instead of get_state_summary() when working in a registered
+    domain repo — returns only what is relevant to the specified domain,
+    typically 80-90% fewer tokens than the full summary.
+
+    Args:
+        domain_slug: the domain slug, e.g. "railiance", "markitect"
+
+    Returns: topic, active workstreams, open blocking decisions for this
+    topic, 5 most recent progress events, and repo SBOM status for this domain.
+    """
+    topics = _get("/topics")
+    topic = next((t for t in topics if t.get("domain_slug") == domain_slug), None)
+    if not topic:
+        return json.dumps({"error": f"No topic found for domain '{domain_slug}'"})
+
+    topic_id = topic["id"]
+
+    workstreams = _get("/workstreams", {"topic_id": topic_id, "status": "active"})
+    blocking = _get("/decisions", {"decision_type": "pending", "topic_id": topic_id})
+    recent = _get("/progress", {"topic_id": topic_id, "limit": 5})
+    repos = _get("/repos", {"domain": domain_slug})
+
+    return json.dumps({
+        "domain": domain_slug,
+        "topic_id": topic_id,
+        "topic_title": topic["title"],
+        "workstreams": workstreams,
+        "blocking_decisions": blocking,
+        "recent_progress": recent,
+        "repos": [{"slug": r["slug"], "last_sbom_at": r.get("last_sbom_at")} for r in repos],
+    }, indent=2)
+
+
 @mcp.tool()
 def get_topic(slug: str) -> str:
     """Return a topic (with workstreams) by slug, plus its recent progress events."""