subagent_generate: accept JSON as arg; remove all escaping from prompt examples

prompts/tools-subagent.md

@@ -6,13 +6,22 @@ Spawn independent subagent sessions with JIT-generated identities via `echo JSON
 
 **Calling convention**:
 ```
-pipe: stages=[{code: "echo '{\"role\": \"ROLE\", \"shared_task\": \"TASK_DESC\"}'"}, {tool: "subagent_generate"}, {tool: "subagent_spawn", args: ["-name", "NAME", "TASK"]}]
+pipe: stages=[
+  {tool: "subagent_generate", args: [{"role": "ROLE", "shared_task": "TASK_DESC"}]},
+  {tool: "subagent_spawn", args: ["-name", "NAME", "TASK"]}
+]
 ```
 
 Parallel — issue separate pipe calls for each subagent (subagent_spawn returns immediately, so sequential dispatch is fine):
 ```
-pipe: stages=[{code: "echo '{\"role\": \"ROLE_A\", \"roster\": \"a (you), b\", \"shared_task\": \"X\"}'"}, {tool: "subagent_generate"}, {tool: "subagent_spawn", args: ["-name", "a", "TASK_A"]}]
-pipe: stages=[{code: "echo '{\"role\": \"ROLE_B\", \"roster\": \"a, b (you)\", \"shared_task\": \"X\"}'"}, {tool: "subagent_generate"}, {tool: "subagent_spawn", args: ["-name", "b", "TASK_B"]}]
+pipe: stages=[
+  {tool: "subagent_generate", args: [{"role": "ROLE_A", "roster": "a (you), b", "shared_task": "X"}]},
+  {tool: "subagent_spawn", args: ["-name", "a", "TASK_A"]}
+]
+pipe: stages=[
+  {tool: "subagent_generate", args: [{"role": "ROLE_B", "roster": "a, b (you)", "shared_task": "X"}]},
+  {tool: "subagent_spawn", args: ["-name", "b", "TASK_B"]}
+]
 ```
 
 **Returns**: `subagent_generate` returns the agent config name. `subagent_spawn` returns the session path (e.g. `s/my-session`).

tools/subagent_generate

@@ -14,7 +14,11 @@ if [ -t 0 ] && [ $# -eq 0 ]; then
     exit 0
 fi
 
-cfg=$(cat) || die "failed to read stdin"
+if [ $# -gt 0 ]; then
+    cfg="$1"
+else
+    cfg=$(cat) || die "failed to read stdin"
+fi
 [ -z "$cfg" ] && die "empty input"
 
 name=$(printf '%s' "$cfg" | python3 -c "