S00-012: Record blocked task and remediation handoff

Changes: Preserve incomplete work and blocker metadata for S00-012. This checkpoint does not mark implementation accepted.
Validation: Session ended with status blocked.
Blocker: Required automation suite fails in existing all-prompts budget test: S21-006 exceeds even a diagnostic 90000-character prompt limit. This blocks verification of the stdin remediation; production limit remains 80000.

Task: S00-012
Agent: manual-ci-cd-stdin-20261015b
Role: ci-cd
Action: escalate
Status: blocked

Task title: Fix runner prompt delivery after partial stdin timeout

scripts/runner.py

@@ -206,21 +206,21 @@ The coordinator commits each transition before launching another agent.
         failure = None
         stopped_after_transition = False
         returncode = None
-        with (session / 'output.log').open('wb') as output:
+        # Give the child a complete UTF-8 stream ending at EOF. Feeding a pipe
+        # through short communicate() retries can strand a partially sent prompt
+        # when the child is slow to start reading. The prompt is already persisted.
+        with prompt_path.open('rb') as prompt_input, (session / 'output.log').open('wb') as output:
             try:
-                process = subprocess.Popen(argv, cwd=self.root, env=env, stdin=subprocess.PIPE,
+                process = subprocess.Popen(argv, cwd=self.root, env=env, stdin=prompt_input,
                                            stdout=output, stderr=subprocess.STDOUT, start_new_session=(os.name == 'posix'))
                 deadline = time.monotonic() + self.timeout
-                first_communicate = True
                 stopped_after_transition = False
                 try:
                     while True:
                         try:
-                            process.communicate(prompt.encode('utf-8') if first_communicate else None,
-                                                timeout=min(.2, max(.01, deadline - time.monotonic())))
+                            process.wait(timeout=min(.2, max(.01, deadline - time.monotonic())))
                             break
                         except subprocess.TimeoutExpired:
-                            first_communicate = False
                             try:
                                 live = self.tracker.state()
                                 current = live['tasks'][item['id']]
@@ -231,17 +231,17 @@ The coordinator commits each transition before launching another agent.
                                 blocker_exit = executor_exit = reviewer_exit = False
                             if blocker_exit or executor_exit or reviewer_exit:
                                 stop_process_tree(process)
-                                process.communicate()
+                                process.wait()
                                 stopped_after_transition = True
                                 break
                             if time.monotonic() >= deadline:
                                 stop_process_tree(process)
-                                process.communicate()
+                                process.wait()
                                 failure = 'Agent timed out; process terminated before escalation'
                                 break
                 except BaseException:
                     stop_process_tree(process)
-                    process.communicate()
+                    process.wait()
                     raise
                 finally:
                     # Even an exited harness must not leave source-writing descendants alive.