fix: 9.5-code reachable items (parse unification, goal branch, fail-open logs, M7 test, fingerprint)

Parse: agent_tools/tool_approvals column parsing unified into agent_def shared helpers (parse None = undeclared; server keeps None, worker maps to []/{} at _task_from_row, dispatch behaviour unchanged). Goal: _apply_resume_preamble goal branch proven reachable via _goal_convert_queued on a queued auto-resumed task, kept with a test and corrected docs (not dead code). Fail-open: audit write failure and preamble lookup failure now log warnings (best-effort unchanged); partial policy failure and shadow parse drops documented as known gaps. M7: shadow endpoint explicit_capability_failures_truncated test. Fingerprint: RoutingPolicy.content_fingerprint (canonical JSON sha256, version excluded) surfaced only in dry-run output, never in audit rows. M1 wording: legacy gap stated as fixed stock persisting per-task until completion, not converging.

cloud/models.py

@@ -414,6 +414,37 @@ class RoutingPolicy:
         """Effective mode for one tier: explicit tier mode, else global."""
         return self.tier_modes.get(tier, self.mode)
 
+    def content_fingerprint(self):
+        """Content fingerprint of the parsed policy (9.5-code).
+
+        sha256 of the canonical JSON (sorted keys, tight
+        separators) over the fields that decide routing and
+        capability checks: mode, tiers, tier_modes,
+        tier_requirements, model_capabilities, fallback_models.
+        The operator-declared ``version`` string is deliberately
+        NOT part of the fingerprint: it is hand-filled and may
+        stay unchanged while requires change, so it cannot be
+        trusted to detect a change. The fingerprint is only for
+        dry-run output and debugging; it is never written into
+        worker_routing audit rows, so the shadow parser (which
+        splits audit detail on whitespace) is untouched.
+        Two policies that parse to the same effective fields
+        share a fingerprint even if their raw documents differ
+        in key order or in dropped/invalid tiers.
+        """
+        import hashlib
+        payload = {
+            "mode": self.mode,
+            "tiers": self.tiers,
+            "tier_modes": self.tier_modes,
+            "tier_requirements": self.tier_requirements,
+            "model_capabilities": self.model_capabilities,
+            "fallback_models": self.fallback_models,
+        }
+        canonical = json.dumps(payload, sort_keys=True,
+                               separators=(",", ":"))
+        return hashlib.sha256(canonical.encode("utf-8")).hexdigest()
+
 
 def parse_routing_policy(data):
     """Parse one policy document, or return None when it is unusable.
@@ -538,6 +569,15 @@ def load_routing_policy(env=None):
                 " worker default", "; ".join(failures))
         return None
     if failures:
+        # 9.5-code fail-open 清单：部分失败是 fail-open——
+        # 可用来源继续生效、失败来源被丢弃，派发不炸。
+        # 可见性只有这条警告日志：load_routing_policy 仍
+        # 返回一个 policy，worker 的
+        # routing_policy_load_failed 标志只在全失败时
+        # 置位，所以部分失败不会产生任务级审计；运营方
+        # 只能从 worker 日志看到它，不能从审计里查到。
+        # 这是已知缺口（非静默，但不如全失败可见），
+        # 改动返回类型会牵动全部调用方，本轨不改。
         _logger.warning(
             "routing policy partially failed (%s); continuing with"
             " the usable source", "; ".join(failures))