ultra-builder-pro-cli 0.26.2 → 3.0.2

package/skills/ultra-review/scripts/review_wait.py

@@ -1,29 +1,50 @@
 #!/usr/bin/env python3
-"""Wait for and validate Ultra review v2 artifacts.
+"""Wait for and validate Ultra review artifacts.
 
 Usage:
+    python3 review_wait.py <session_path> packet --packet-digest <sha256>
     python3 review_wait.py <session_path> agents --packet-digest <sha256> <artifact-stem> [<artifact-stem> ...]
     python3 review_wait.py <session_path> summary --packet-digest <sha256>
+    python3 review_wait.py <session_path> summary --packet-digest <sha256> --summary-snapshot-digest <sha256>
+
+New sessions require admitted v4. Add ``--legacy-v4`` immediately after the packet
+digest only when reading an immutable pre-admission v4 session, or ``--legacy-v3`` for
+an immutable historical v3 session; neither flag writes or upgrades artifacts.
 
 The agents mode validates the exact named JSON files. The summary mode validates the
 two-axis verdict contract. Timeout and poll intervals can be shortened in tests with
 UBP_REVIEW_WAIT_TIMEOUT and UBP_REVIEW_WAIT_POLL.
 """
 
+import hashlib
 import json
 import os
 import re
+import selectors
+import shutil
+import stat
+import subprocess
 import sys
+import tempfile
 import time
-from pathlib import Path
+from datetime import datetime
+from pathlib import Path, PurePosixPath
 
-FINDINGS_SCHEMA = "ultra-review-findings-v2"
-SUMMARY_SCHEMA = "ultra-review-summary-v2"
+FINDINGS_SCHEMA = "ultra-review-findings-v4"
+SUMMARY_SCHEMA = "ultra-review-summary-v4"
+LEGACY_FINDINGS_SCHEMA = "ultra-review-findings-v3"
+LEGACY_SUMMARY_SCHEMA = "ultra-review-summary-v3"
+ADMISSION_SCHEMA_V1 = "ultra-review-admission-v1"
+ADMISSION_SCHEMA_V2 = "ultra-review-admission-v2"
+ADMISSION_CONTRACT_V1 = "ultra-review-admission-required-v1"
+ADMISSION_CONTRACT_V2 = "ultra-review-admission-required-v2"
+STRICT_ADMISSION_CONTRACTS = {ADMISSION_CONTRACT_V1, ADMISSION_CONTRACT_V2}
 AXES = {"spec_fidelity", "engineering_standards"}
 SEVERITIES = {"P0", "P1", "P2", "P3"}
 AXIS_VERDICTS = {"PASS", "FAIL", "INCOMPLETE"}
 OVERALL_VERDICTS = {"APPROVE", "REQUEST_CHANGES", "INCOMPLETE"}
 REVIEW_MODES = {"task", "change", "plan"}
+EXECUTION_MODES = {"isolated", "sequential-shared-context"}
 SELECTION_STATUSES = {"selected", "skipped"}
 REVIEW_WORKERS = {
     "review-spec",
@@ -33,11 +54,83 @@ REVIEW_WORKERS = {
     "review-design",
     "review-comments",
 }
+CANONICAL_WORKERS = (
+    ("review-spec", "spec_fidelity", "skills/ultra-review/references/spec.md"),
+    ("review-code", "engineering_standards", "skills/ultra-review/references/code.md"),
+    ("review-tests", "engineering_standards", "skills/ultra-review/references/tests.md"),
+    ("review-errors", "engineering_standards", "skills/ultra-review/references/errors.md"),
+    ("review-design", "engineering_standards", "skills/ultra-review/references/design.md"),
+    ("review-comments", "engineering_standards", "skills/ultra-review/references/comments.md"),
+)
+CANONICAL_WORKER_BY_AGENT = {
+    agent: {"axis": axis, "lens": lens, "order": order}
+    for order, (agent, axis, lens) in enumerate(CANONICAL_WORKERS)
+}
+PACKET_REQUIRED_FIELDS = {
+    "$schema",
+    "session",
+    "mode",
+    "created_at",
+    "head",
+    "range",
+    "change_id",
+    "task_ids",
+    "acceptance",
+    "public_seams",
+    "north_star_trace",
+    "context_files",
+    "workers",
+    "diff_files",
+    "output_directory",
+}
+TRACE_FIELDS = {
+    "path",
+    "first_principles",
+    "serves",
+    "touches",
+    "north_star_revision",
+    "north_star_digest",
+}
+CONTEXT_FIELDS = {"path", "sha256"}
+WORKER_FIELDS = {"agent", "axis", "lens", "output"}
+REVIEW_HISTORY_FIELDS = {
+    "parent_summary_ref",
+    "parent_summary_digest",
+    "unresolved_blocking_ids",
+}
+FOREIGN_REVIEW_SESSION = re.compile(
+    r"\.ultra/reviews/([A-Za-z0-9][A-Za-z0-9._-]*)/"
+)
 ARTIFACT_STEM = re.compile(r"^[a-z][a-z0-9-]*$")
 FINDING_FIELDS = {
     "id", "axis", "severity", "category", "title", "file", "line", "trigger",
     "impact", "evidence", "suggestion",
 }
+MAX_JSON_BYTES = 8 * 1024 * 1024
+READ_CHUNK_BYTES = 64 * 1024
+MAX_VALIDATOR_REPORT_BYTES = 1024 * 1024
+VALIDATOR_TIMEOUT_SECONDS = 5
+ADMISSION_FIELDS = {
+    "$schema",
+    "version",
+    "session",
+    "packet_digest",
+    "head",
+    "observations",
+    "north_star_report",
+    "subject_digest",
+}
+ADMISSION_OBSERVATION_FIELDS_V1 = {"role", "path", "sha256"}
+ADMISSION_OBSERVATION_FIELDS_V2 = {"role", "path", "sha256", "byte_length"}
+PACKET_SUBJECT_OBSERVATION_FIELDS = {"role", "path", "sha256", "byte_length"}
+NORTH_STAR_REPORT_RECEIPT_FIELDS = {
+    "schema",
+    "sha256",
+    "input_sha256",
+    "input_byte_length",
+    "source_observations",
+}
+NORTH_STAR_SOURCE_OBSERVATION_FIELDS = {"role", "path", "sha256", "byte_length"}
 
 
 def env_seconds(name: str, default: float) -> float:
@@ -52,16 +145,229 @@ DEFAULT_TIMEOUT = env_seconds("UBP_REVIEW_WAIT_TIMEOUT", 300.0)
 POLL_INTERVAL = env_seconds("UBP_REVIEW_WAIT_POLL", 2.0)
 
 
-def read_json(path: Path):
+def decode_json_bytes(data_bytes: bytes, label: str):
     try:
-        data = json.loads(path.read_text(encoding="utf-8"))
-    except (OSError, json.JSONDecodeError) as error:
-        return None, f"unreadable JSON: {error}"
+        data = json.loads(data_bytes.decode("utf-8"))
+    except UnicodeDecodeError as error:
+        return None, f"{label} is unreadable UTF-8 JSON: {error}"
+    except json.JSONDecodeError as error:
+        return None, f"{label} is unreadable JSON: {error}"
     if not isinstance(data, dict):
-        return None, "top-level value must be an object"
+        return None, f"{label} top-level value must be an object"
     return data, None
 
 
+def _stat_identity(value):
+    return (
+        value.st_dev,
+        value.st_ino,
+        value.st_mode,
+        value.st_size,
+        value.st_mtime_ns,
+        value.st_ctime_ns,
+    )
+
+
+def _component_identity(value):
+    return (value.st_dev, value.st_ino, value.st_mode)
+
+
+def _verify_fresh_repo_walk(
+    root: Path,
+    segments,
+    directory_identities,
+    file_identity,
+    label: str,
+):
+    """Rewalk from the current root and reject any swapped path component."""
+    directory_flags = os.O_RDONLY
+    directory_flags |= getattr(os, "O_DIRECTORY", 0)
+    directory_flags |= getattr(os, "O_NOFOLLOW", 0)
+    descriptors = []
+    try:
+        root_path = os.stat(root, follow_symlinks=False)
+        descriptor = os.open(root, directory_flags)
+        descriptors.append(descriptor)
+        root_fd = os.fstat(descriptor)
+        if (
+            _component_identity(root_path) != directory_identities[0]
+            or _component_identity(root_fd) != directory_identities[0]
+        ):
+            return f"{label} repository root changed after its stable byte snapshot"
+
+        for index, segment in enumerate(segments[:-1], start=1):
+            component = os.stat(segment, dir_fd=descriptor, follow_symlinks=False)
+            if stat.S_ISLNK(component.st_mode) or not stat.S_ISDIR(component.st_mode):
+                return f"{label} path component changed after its stable byte snapshot"
+            next_descriptor = os.open(segment, directory_flags, dir_fd=descriptor)
+            descriptors.append(next_descriptor)
+            opened = os.fstat(next_descriptor)
+            if (
+                _component_identity(component) != directory_identities[index]
+                or _component_identity(opened) != directory_identities[index]
+            ):
+                return (
+                    f"{label} path component {'/'.join(segments[:index])} changed "
+                    "after its stable byte snapshot"
+                )
+            descriptor = next_descriptor
+
+        current_file = os.stat(
+            segments[-1],
+            dir_fd=descriptor,
+            follow_symlinks=False,
+        )
+        if (
+            not stat.S_ISREG(current_file.st_mode)
+            or _stat_identity(current_file) != file_identity
+        ):
+            return f"{label} changed after its stable byte snapshot"
+    except (FileNotFoundError, NotADirectoryError):
+        return f"{label} path changed after its stable byte snapshot"
+    except OSError as error:
+        return f"{label} fresh path identity cannot be verified: {error}"
+    finally:
+        for descriptor in reversed(descriptors):
+            os.close(descriptor)
+    return None
+
+
+def stable_read_repo_bytes(
+    repo_root: Path,
+    repo_path: str,
+    label: str,
+    max_bytes: int = MAX_JSON_BYTES,
+):
+    """Read one bounded file through a no-symlink repository descriptor walk."""
+    normalized, path_error = validate_repo_path(repo_path, repo_root, label)
+    if path_error:
+        return None, path_error, False
+    segments = normalized.split("/")
+    directory_flags = os.O_RDONLY
+    directory_flags |= getattr(os, "O_DIRECTORY", 0)
+    directory_flags |= getattr(os, "O_NOFOLLOW", 0)
+    file_flags = os.O_RDONLY
+    file_flags |= getattr(os, "O_NONBLOCK", 0)
+    file_flags |= getattr(os, "O_NOFOLLOW", 0)
+    directory_descriptors = []
+    descriptor = None
+    directory_identities = []
+    try:
+        root = repo_root.resolve(strict=True)
+        root_path = os.stat(root, follow_symlinks=False)
+        directory_descriptor = os.open(root, directory_flags)
+        directory_descriptors.append(directory_descriptor)
+        root_fd = os.fstat(directory_descriptor)
+        if _component_identity(root_path) != _component_identity(root_fd):
+            return None, f"{label} repository root changed before its stable byte snapshot", False
+        directory_identities.append(_component_identity(root_fd))
+        for segment in segments[:-1]:
+            component = os.stat(
+                segment,
+                dir_fd=directory_descriptor,
+                follow_symlinks=False,
+            )
+            if stat.S_ISLNK(component.st_mode):
+                return None, f"{label} path components must not be symlinks", False
+            if not stat.S_ISDIR(component.st_mode):
+                return None, f"{label} path component is not a directory", False
+            next_descriptor = os.open(
+                segment,
+                directory_flags,
+                dir_fd=directory_descriptor,
+            )
+            directory_descriptors.append(next_descriptor)
+            opened_component = os.fstat(next_descriptor)
+            if _component_identity(component) != _component_identity(opened_component):
+                return None, (
+                    f"{label} path component changed before its stable byte snapshot"
+                ), False
+            directory_identities.append(_component_identity(opened_component))
+            directory_descriptor = next_descriptor
+
+        filename = segments[-1]
+        before_path = os.stat(
+            filename,
+            dir_fd=directory_descriptor,
+            follow_symlinks=False,
+        )
+        if stat.S_ISLNK(before_path.st_mode):
+            return None, (
+                f"{label} must be a stable repository-contained regular file, not a symlink"
+            ), False
+        if not stat.S_ISREG(before_path.st_mode):
+            return None, f"{label} must be a stable repository-contained regular file", False
+        if before_path.st_size > max_bytes:
+            return None, f"{label} exceeds the {max_bytes}-byte artifact limit", False
+
+        descriptor = os.open(filename, file_flags, dir_fd=directory_descriptor)
+        before_fd = os.fstat(descriptor)
+        if not stat.S_ISREG(before_fd.st_mode):
+            return None, f"{label} must be a stable repository-contained regular file", False
+        if _stat_identity(before_path) != _stat_identity(before_fd):
+            return None, f"{label} changed before its stable byte snapshot could be read", False
+
+        chunks = []
+        observed = 0
+        while True:
+            chunk = os.read(descriptor, min(READ_CHUNK_BYTES, max_bytes + 1 - observed))
+            if not chunk:
+                break
+            chunks.append(chunk)
+            observed += len(chunk)
+            if observed > max_bytes:
+                return None, f"{label} exceeds the {max_bytes}-byte artifact limit", False
+        after_fd = os.fstat(descriptor)
+        after_path = os.stat(
+            filename,
+            dir_fd=directory_descriptor,
+            follow_symlinks=False,
+        )
+    except FileNotFoundError:
+        return None, f"{label} is missing", True
+    except OSError as error:
+        return None, f"{label} cannot be read as a stable regular file: {error}", False
+    finally:
+        if descriptor is not None:
+            os.close(descriptor)
+        for directory_descriptor in reversed(directory_descriptors):
+            os.close(directory_descriptor)
+
+    if (
+        _stat_identity(before_fd) != _stat_identity(after_fd)
+        or _stat_identity(after_fd) != _stat_identity(after_path)
+    ):
+        return None, f"{label} changed while its stable byte snapshot was read", False
+
+    data_bytes = b"".join(chunks)
+    if len(data_bytes) != after_fd.st_size:
+        return None, f"{label} size changed while its stable byte snapshot was read", False
+    rewalk_error = _verify_fresh_repo_walk(
+        root,
+        segments,
+        directory_identities,
+        _stat_identity(after_fd),
+        label,
+    )
+    if rewalk_error:
+        return None, rewalk_error, False
+    return {
+        "bytes": data_bytes,
+        "sha256": hashlib.sha256(data_bytes).hexdigest(),
+        "identity": _stat_identity(after_fd),
+        "component_identities": tuple(directory_identities),
+        "path": normalized,
+    }, None, False
+
+
+def stable_read_repo_json(repo_root: Path, repo_path: str, label: str):
+    snapshot, error, missing = stable_read_repo_bytes(repo_root, repo_path, label)
+    if error:
+        return None, snapshot, error, missing
+    data, error = decode_json_bytes(snapshot["bytes"], label)
+    return data, snapshot, error, False
+
+
 def nonempty_string(value) -> bool:
     return isinstance(value, str) and bool(value.strip())
 
@@ -70,7 +376,1612 @@ def digest_string(value) -> bool:
     return isinstance(value, str) and re.fullmatch(r"[0-9a-f]{64}", value) is not None
 
 
-def validate_finding(finding, artifact_axis: str, seen_ids):
+def git_blob_digest_string(value) -> bool:
+    return isinstance(value, str) and re.fullmatch(r"[0-9a-f]{40}", value) is not None
+
+
+def timestamp_string(value) -> bool:
+    if not isinstance(value, str) or re.fullmatch(
+        r"\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}(?:\.\d+)?(?:Z|[+-]\d{2}:\d{2})",
+        value,
+    ) is None:
+        return False
+    try:
+        parsed = datetime.fromisoformat(value[:-1] + "+00:00" if value.endswith("Z") else value)
+    except ValueError:
+        return False
+    return parsed.tzinfo is not None
+
+
+def canonical_repo_root(session_path: Path):
+    resolved_session = session_path.resolve()
+    if (
+        resolved_session.parent.name != "reviews"
+        or resolved_session.parent.parent.name != ".ultra"
+    ):
+        return None, (
+            "session path must resolve to the canonical repository location "
+            "<repo>/.ultra/reviews/<session>"
+        )
+    return resolved_session.parent.parent.parent, None
+
+
+def validate_repo_path(value, repo_root: Path, field: str, allow_fragment=False):
+    if not nonempty_string(value):
+        return None, f"{field} must be a non-empty normalized repository-relative path"
+
+    path_value = value
+    if allow_fragment:
+        if value.count("#") != 1:
+            return None, f"{field} must contain one non-empty fragment"
+        path_value, fragment = value.split("#", 1)
+        if re.fullmatch(r"[A-Za-z0-9][A-Za-z0-9._-]*", fragment) is None:
+            return None, f"{field} has an invalid fragment"
+    elif "#" in value:
+        return None, f"{field} must not contain a fragment"
+
+    if (
+        "\\" in path_value
+        or "\x00" in path_value
+        or path_value.startswith("/")
+        or re.match(r"^[A-Za-z]:", path_value)
+    ):
+        return None, f"{field} must be a normalized repository-relative POSIX path"
+    segments = path_value.split("/")
+    if any(segment in {"", ".", ".."} for segment in segments):
+        return None, f"{field} must not contain empty, dot, or parent path segments"
+    if PurePosixPath(path_value).as_posix() != path_value:
+        return None, f"{field} must be normalized"
+
+    try:
+        resolved_root = repo_root.resolve()
+        resolved_target = (resolved_root / Path(*segments)).resolve(strict=False)
+    except (OSError, RuntimeError) as error:
+        return None, f"{field} cannot be resolved safely: {error}"
+    try:
+        resolved_target.relative_to(resolved_root)
+    except ValueError:
+        return None, f"{field} escapes the repository root"
+
+    # Path.resolve(strict=False) stopped raising on symlink loops in Python
+    # 3.13, so a looping or later-swapped component can survive resolution
+    # while the stable-read descriptor walks would still refuse it. Refuse
+    # statable symlink or non-directory components here too; components that
+    # merely do not exist yet stay valid at validation time.
+    probe = resolved_root
+    for segment in segments[:-1]:
+        probe = probe / segment
+        try:
+            component_stat = os.stat(probe, follow_symlinks=False)
+        except FileNotFoundError:
+            continue
+        except NotADirectoryError:
+            return None, f"{field} must traverse ordinary directories"
+        if stat.S_ISLNK(component_stat.st_mode) or not stat.S_ISDIR(component_stat.st_mode):
+            return None, f"{field} must traverse ordinary non-symlink directories"
+    return path_value, None
+
+
+def validate_string_array(value, field: str, require_nonempty=False):
+    if not isinstance(value, list) or not all(nonempty_string(item) for item in value):
+        return f"{field} must be a string array"
+    if require_nonempty and not value:
+        return f"{field} must not be empty"
+    if len(value) != len(set(value)):
+        return f"{field} must not contain duplicates"
+    return None
+
+
+def canonical_context_digest(context_files) -> str:
+    if len(context_files) == 1:
+        return context_files[0]["sha256"]
+    canonical = sorted(
+        (
+            {"path": item["path"], "sha256": item["sha256"]}
+            for item in context_files
+        ),
+        key=lambda item: item["path"].encode("utf-8"),
+    )
+    payload = json.dumps(
+        canonical,
+        ensure_ascii=False,
+        separators=(",", ":"),
+        sort_keys=True,
+    ).encode("utf-8")
+    return hashlib.sha256(payload).hexdigest()
+
+
+def _trace_arrays(value):
+    if not isinstance(value, dict):
+        return None
+    result = {}
+    patterns = {
+        "first_principles": r"FP-[A-Za-z0-9][A-Za-z0-9._-]*",
+        "serves": r"NS-[A-Za-z0-9][A-Za-z0-9._-]*",
+        "touches": r"HC-[A-Za-z0-9][A-Za-z0-9._-]*",
+    }
+    for field, pattern in patterns.items():
+        values = value.get(field)
+        if not isinstance(values, list) or not all(
+            isinstance(item, str) and re.fullmatch(pattern, item) for item in values
+        ) or len(values) != len(set(values)):
+            return None
+        result[field] = values
+    return result
+
+
+def _terminate_and_reap(process):
+    if process.poll() is None:
+        process.terminate()
+        try:
+            process.wait(timeout=0.25)
+        except subprocess.TimeoutExpired:
+            process.kill()
+            process.wait()
+    else:
+        process.wait()
+
+
+def _bounded_process_output(argv, input_bytes: bytes, timeout: float, max_output_bytes: int):
+    try:
+        process = subprocess.Popen(
+            argv,
+            stdin=subprocess.PIPE,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,
+            bufsize=0,
+        )
+    except OSError as error:
+        return None, None, None, f"canonical North Star validator could not run: {error}"
+
+    selector = selectors.DefaultSelector()
+    buffers = {"stdout": bytearray(), "stderr": bytearray()}
+    input_view = memoryview(input_bytes)
+    input_offset = 0
+    combined_output = 0
+    failure = None
+    returncode = None
+
+    def stop_watching(stream):
+        try:
+            selector.unregister(stream)
+        except (KeyError, ValueError):
+            pass
+        try:
+            stream.close()
+        except OSError:
+            pass
+
+    try:
+        for stream, role, events in [
+            (process.stdout, "stdout", selectors.EVENT_READ),
+            (process.stderr, "stderr", selectors.EVENT_READ),
+            (process.stdin, "stdin", selectors.EVENT_WRITE),
+        ]:
+            os.set_blocking(stream.fileno(), False)
+            selector.register(stream, events, role)
+        if not input_bytes:
+            stop_watching(process.stdin)
+
+        deadline = time.monotonic() + timeout
+        while selector.get_map():
+            remaining = deadline - time.monotonic()
+            if remaining <= 0:
+                failure = "canonical North Star validator timed out"
+                break
+            events = selector.select(min(0.05, remaining))
+            for key, _mask in events:
+                stream = key.fileobj
+                role = key.data
+                if role == "stdin":
+                    try:
+                        written = os.write(
+                            stream.fileno(),
+                            input_view[input_offset:input_offset + READ_CHUNK_BYTES],
+                        )
+                    except BlockingIOError:
+                        continue
+                    except BrokenPipeError:
+                        stop_watching(stream)
+                        continue
+                    input_offset += written
+                    if input_offset == len(input_view):
+                        stop_watching(stream)
+                    continue
+
+                try:
+                    chunk = os.read(stream.fileno(), READ_CHUNK_BYTES)
+                except BlockingIOError:
+                    continue
+                if not chunk:
+                    stop_watching(stream)
+                    continue
+                combined_output += len(chunk)
+                if combined_output > max_output_bytes:
+                    failure = (
+                        "canonical North Star validator output exceeded its bounded "
+                        "combined report limit"
+                    )
+                    break
+                buffers[role].extend(chunk)
+            if failure:
+                break
+
+        if failure:
+            _terminate_and_reap(process)
+        else:
+            remaining = max(0.0, deadline - time.monotonic())
+            try:
+                returncode = process.wait(timeout=remaining)
+            except subprocess.TimeoutExpired:
+                failure = "canonical North Star validator timed out"
+                _terminate_and_reap(process)
+    finally:
+        selector.close()
+        for stream in (process.stdin, process.stdout, process.stderr):
+            if stream is not None and not stream.closed:
+                try:
+                    stream.close()
+                except OSError:
+                    pass
+        if process.poll() is None:
+            _terminate_and_reap(process)
+
+    return returncode, bytes(buffers["stdout"]), bytes(buffers["stderr"]), failure
+
+
+def validate_north_star_snapshot(repo_root: Path, snapshot):
+    node = shutil.which("node")
+    if node is None:
+        return None, "canonical North Star validation requires the installed node runtime"
+    validator = (
+        Path(__file__).resolve().parents[2]
+        / "ultra-research"
+        / "scripts"
+        / "validate_north_star.cjs"
+    )
+    if not validator.is_file():
+        return None, f"canonical North Star validator is missing: {validator}"
+    canonical_path = repo_root.resolve() / ".ultra" / "north-star.md"
+    returncode, stdout, stderr, process_error = _bounded_process_output(
+        [node, str(validator), "--stdin", "--path", str(canonical_path)],
+        snapshot["bytes"],
+        VALIDATOR_TIMEOUT_SECONDS,
+        MAX_VALIDATOR_REPORT_BYTES,
+    )
+    if process_error:
+        return None, process_error
+    report, report_error = decode_json_bytes(
+        stdout,
+        "canonical North Star validator report",
+    )
+    if report_error:
+        detail = stderr.decode("utf-8", errors="replace").strip()
+        return None, f"{report_error}{': ' + detail if detail else ''}"
+    input_receipt = report.get("input")
+    expected_input = {
+        "path": str(canonical_path),
+        "byte_length": len(snapshot["bytes"]),
+        "sha256": snapshot["sha256"],
+    }
+    if input_receipt != expected_input:
+        return None, (
+            "canonical North Star validator report is not bound to the exact stdin "
+            "snapshot bytes and canonical path"
+        )
+    binding = report.get("acceptance_binding")
+    source_observations = report.get("source_observations")
+    expected_sources = None
+    if isinstance(binding, dict):
+        source = binding.get("source")
+        accepted_snapshot = binding.get("snapshot")
+        if isinstance(source, str) and "#" in source and isinstance(accepted_snapshot, str):
+            expected_sources = [
+                ("decision", source.split("#", 1)[0]),
+                ("snapshot", accepted_snapshot),
+            ]
+    if expected_sources is None or not isinstance(source_observations, list):
+        return None, "canonical North Star validator report lacks exact source observations"
+    observed_sources = []
+    for index, observation in enumerate(source_observations):
+        if (
+            not isinstance(observation, dict)
+            or set(observation) != NORTH_STAR_SOURCE_OBSERVATION_FIELDS
+            or observation.get("role") not in {"decision", "snapshot"}
+            or not digest_string(observation.get("sha256"))
+            or not isinstance(observation.get("byte_length"), int)
+            or isinstance(observation.get("byte_length"), bool)
+            or observation["byte_length"] < 0
+        ):
+            return None, (
+                "canonical North Star validator report has an invalid "
+                f"source_observations[{index}] receipt"
+            )
+        normalized, path_error = validate_repo_path(
+            observation.get("path"),
+            repo_root,
+            f"canonical North Star validator source_observations[{index}].path",
+        )
+        if path_error:
+            return None, path_error
+        observed_sources.append((observation["role"], normalized))
+    if observed_sources != expected_sources:
+        return None, (
+            "canonical North Star validator report source observations do not exactly "
+            "match its decision and accepted snapshot binding"
+        )
+    if returncode != 0 or report.get("valid") is not True:
+        codes = [
+            item.get("code")
+            for item in report.get("diagnostics", [])
+            if isinstance(item, dict) and nonempty_string(item.get("code"))
+        ]
+        return None, (
+            "canonical North Star validator rejected the current snapshot"
+            + (f": {', '.join(codes)}" if codes else "")
+        )
+    return {
+        "data": report,
+        "sha256": hashlib.sha256(stdout).hexdigest(),
+    }, None
+
+
+def current_git_head(repo_root: Path):
+    try:
+        completed = subprocess.run(
+            ["git", "-C", str(repo_root), "rev-parse", "--verify", "HEAD"],
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,
+            timeout=VALIDATOR_TIMEOUT_SECONDS,
+            check=False,
+        )
+    except subprocess.TimeoutExpired:
+        return None, "current Git HEAD observation timed out"
+    except OSError as error:
+        return None, f"current Git HEAD cannot be observed: {error}"
+    if completed.returncode != 0:
+        detail = completed.stderr.decode("utf-8", errors="replace").strip()
+        return None, "current Git HEAD cannot be observed" + (f": {detail}" if detail else "")
+    try:
+        head = completed.stdout.decode("ascii").strip()
+    except UnicodeDecodeError:
+        return None, "current Git HEAD is not ASCII"
+    if not git_blob_digest_string(head):
+        return None, f"current Git HEAD is not a lowercase 40-hex commit digest: {head!r}"
+    return head, None
+
+
+def acceptance_source_paths(packet):
+    return sorted(
+        {
+            acceptance.partition(": ")[0].split("#", 1)[0]
+            for acceptance in packet["acceptance"]
+        },
+        key=lambda value: value.encode("utf-8"),
+    )
+
+
+def trace_source_path(packet):
+    return packet["north_star_trace"]["path"].split("#", 1)[0]
+
+
+def validate_packet_subject_observations(packet, repo_root: Path):
+    observations = packet.get("subject_observations")
+    acceptance_paths = acceptance_source_paths(packet)
+    expected_count = 1 + len(acceptance_paths) + 2
+    if not isinstance(observations, list) or len(observations) != expected_count:
+        return (
+            "WORKER-PACKET.json subject_observations must be the exact ordered "
+            "Change, acceptance-source, decision, and snapshot observation array"
+        )
+
+    expected_prefix = [
+        ("change", trace_source_path(packet)),
+        *(("acceptance_source", path) for path in acceptance_paths),
+    ]
+    expected_roles = [
+        *(role for role, _path in expected_prefix),
+        "decision",
+        "snapshot",
+    ]
+    observed_roles_and_paths = []
+    for index, observation in enumerate(observations):
+        if (
+            not isinstance(observation, dict)
+            or set(observation) != PACKET_SUBJECT_OBSERVATION_FIELDS
+        ):
+            return (
+                f"WORKER-PACKET.json subject_observations[{index}] must contain "
+                "exactly role, path, sha256, and byte_length"
+            )
+        if observation.get("role") != expected_roles[index]:
+            return (
+                "WORKER-PACKET.json subject_observations must use exact ordered roles: "
+                + ", ".join(expected_roles)
+            )
+        normalized, path_error = validate_repo_path(
+            observation.get("path"),
+            repo_root,
+            f"WORKER-PACKET.json subject_observations[{index}].path",
+        )
+        if path_error:
+            return path_error
+        if not digest_string(observation.get("sha256")):
+            return (
+                f"WORKER-PACKET.json subject_observations[{index}].sha256 must be a "
+                "lowercase SHA-256 digest"
+            )
+        byte_length = observation.get("byte_length")
+        if (
+            not isinstance(byte_length, int)
+            or isinstance(byte_length, bool)
+            or byte_length < 0
+        ):
+            return (
+                f"WORKER-PACKET.json subject_observations[{index}].byte_length must "
+                "be a non-negative integer"
+            )
+        observed_roles_and_paths.append((observation["role"], normalized))
+
+    if observed_roles_and_paths[:len(expected_prefix)] != expected_prefix:
+        return (
+            "WORKER-PACKET.json subject_observations must bind the exact validated "
+            "Change path and ordered acceptance sources"
+        )
+    return None
+
+
+def canonical_admission_subject(value):
+    return {
+        "version": value["version"],
+        "session": value["session"],
+        "packet_digest": value["packet_digest"],
+        "head": value["head"],
+        "observations": value["observations"],
+        "north_star_report": value["north_star_report"],
+    }
+
+
+def canonical_json_digest(value):
+    payload = json.dumps(
+        value,
+        ensure_ascii=False,
+        separators=(",", ":"),
+        sort_keys=True,
+    ).encode("utf-8")
+    return hashlib.sha256(payload).hexdigest()
+
+
+def validate_packet_admission(packet, repo_root: Path, packet_digest: str):
+    observed_head, error = current_git_head(repo_root)
+    if error:
+        return None, error
+    if observed_head != packet["head"]:
+        return None, (
+            f"WORKER-PACKET.json head does not match current Git HEAD {observed_head}; "
+            "create a new review session for the current subject"
+        )
+
+    snapshots = {}
+
+    def subject_snapshot(repo_path, label):
+        if repo_path not in snapshots:
+            snapshot, snapshot_error, _missing = stable_read_repo_bytes(
+                repo_root,
+                repo_path,
+                label,
+            )
+            if snapshot_error:
+                return None, snapshot_error
+            snapshots[repo_path] = snapshot
+        return snapshots[repo_path], None
+
+    observations = []
+    for index, context_file in enumerate(packet["context_files"]):
+        snapshot, error = subject_snapshot(
+            context_file["path"],
+            f"WORKER-PACKET.json context_files[{index}] subject",
+        )
+        if error:
+            return None, error
+        observed = snapshot["sha256"]
+        if observed != context_file["sha256"]:
+            return None, (
+                f"WORKER-PACKET.json context_files[{index}] subject bytes do not match "
+                f"recorded sha256 {context_file['sha256']}; observed {observed}"
+            )
+        observations.append({
+            "role": "context",
+            "path": context_file["path"],
+            "sha256": observed,
+            "byte_length": len(snapshot["bytes"]),
+        })
+
+    north_star_path = ".ultra/north-star.md"
+    north_star, error = subject_snapshot(
+        north_star_path,
+        "WORKER-PACKET.json canonical North Star subject",
+    )
+    if error:
+        return None, error
+    validated_report, error = validate_north_star_snapshot(repo_root, north_star)
+    if error:
+        return None, error
+    report = validated_report["data"]
+    if (
+        report.get("$schema") != "ultra-north-star-validation-v1"
+        or report.get("kind") != "north-star-v2"
+        or report.get("status") != "accepted"
+        or report.get("classification") != "accepted"
+    ):
+        return None, "canonical North Star validator report must describe an accepted v2 authority"
+
+    trace = packet["north_star_trace"]
+    if report.get("revision") != trace["north_star_revision"]:
+        return None, (
+            "north_star_trace.north_star_revision does not match the canonical "
+            f"validator report; observed {report.get('revision')!r}"
+        )
+    binding = report.get("acceptance_binding")
+    if not isinstance(binding, dict):
+        return None, "canonical North Star validator report lacks acceptance binding evidence"
+    if binding.get("git_blob_digest") != trace["north_star_digest"]:
+        return None, (
+            "north_star_trace.north_star_digest does not match the canonical validator "
+            f"report; observed {binding.get('git_blob_digest')!r}"
+        )
+    if binding.get("content_sha256") != north_star["sha256"]:
+        return None, "canonical North Star validator report is not bound to the stable input bytes"
+
+    ids = report.get("ids")
+    if not isinstance(ids, dict) or set(ids) != {"FP", "NS", "HC"}:
+        return None, "canonical North Star validator report has an invalid ID inventory"
+    for field, kind in [
+        ("first_principles", "FP"),
+        ("serves", "NS"),
+        ("touches", "HC"),
+    ]:
+        available = ids.get(kind)
+        if not isinstance(available, list) or not all(nonempty_string(item) for item in available):
+            return None, f"canonical North Star validator report has invalid {kind} IDs"
+        unresolved = [item for item in trace[field] if item not in set(available)]
+        if unresolved:
+            return None, (
+                f"north_star_trace.{field} contains IDs unresolved by the canonical "
+                f"validator report: {', '.join(unresolved)}"
+            )
+
+    observations.append({
+        "role": "north_star",
+        "path": north_star_path,
+        "sha256": north_star["sha256"],
+        "byte_length": len(north_star["bytes"]),
+    })
+
+    packet_subjects = packet["subject_observations"]
+    canonical_sources = packet_subjects[-2:]
+    if report.get("source_observations") != canonical_sources:
+        return None, (
+            "canonical North Star validator decision and snapshot observations do not "
+            "match WORKER-PACKET.json subject_observations"
+        )
+    for index, expected_observation in enumerate(packet_subjects):
+        source, error = subject_snapshot(
+            expected_observation["path"],
+            f"WORKER-PACKET.json subject_observations[{index}] subject",
+        )
+        if error:
+            return None, error
+        observed_length = len(source["bytes"])
+        if (
+            source["sha256"] != expected_observation["sha256"]
+            or observed_length != expected_observation["byte_length"]
+        ):
+            return None, (
+                f"WORKER-PACKET.json subject_observations[{index}] bytes do not match "
+                f"recorded byte_length and sha256 for {expected_observation['path']}"
+            )
+        if expected_observation["role"] in {"change", "acceptance_source"}:
+            observations.append(dict(expected_observation))
+
+    final_head, error = current_git_head(repo_root)
+    if error:
+        return None, error
+    if final_head != observed_head:
+        return None, "current Git HEAD changed during packet admission; retry with a new session"
+    for repo_path, expected in snapshots.items():
+        current, snapshot_error, _missing = stable_read_repo_bytes(
+            repo_root,
+            repo_path,
+            f"WORKER-PACKET.json admitted subject {repo_path}",
+        )
+        if snapshot_error:
+            return None, snapshot_error
+        if (
+            current["bytes"] != expected["bytes"]
+            or len(current["bytes"]) != len(expected["bytes"])
+            or current["sha256"] != expected["sha256"]
+        ):
+            return None, (
+                f"admitted subject {repo_path} bytes, byte length, or sha256 changed "
+                "during packet admission"
+            )
+
+    report_receipt = {
+        "schema": report["$schema"],
+        "sha256": validated_report["sha256"],
+        "input_sha256": north_star["sha256"],
+        "input_byte_length": len(north_star["bytes"]),
+        "source_observations": report["source_observations"],
+    }
+    admission = {
+        "$schema": ADMISSION_SCHEMA_V2,
+        "version": 2,
+        "session": packet["session"],
+        "packet_digest": packet_digest,
+        "head": packet["head"],
+        "observations": observations,
+        "north_star_report": report_receipt,
+    }
+    admission["subject_digest"] = canonical_json_digest(
+        canonical_admission_subject(admission)
+    )
+    return admission, None
+
+
+def cleanup_admission_temporary(
+    session_descriptor: int,
+    temporary: str,
+    publication_descriptor: int,
+):
+    """Remove only the temporary pathname that still names our open inode."""
+    filename = os.path.basename(os.fspath(temporary))
+    try:
+        current = os.lstat(filename, dir_fd=session_descriptor)
+        owned = os.fstat(publication_descriptor)
+    except FileNotFoundError:
+        return True, None
+    except OSError as error:
+        return False, error
+    if (
+        not stat.S_ISREG(current.st_mode)
+        or _stat_identity(current) != _stat_identity(owned)
+    ):
+        return False, (
+            "temporary pathname identity differs from the owned admission file; "
+            "the replacement was retained for manual recovery"
+        )
+    try:
+        os.unlink(filename, dir_fd=session_descriptor)
+    except OSError as error:
+        return False, error
+    return True, None
+
+
+def atomic_write_admission(session_path: Path, admission):
+    target = session_path / "ADMISSION.json"
+    payload = (json.dumps(admission, ensure_ascii=False, indent=2) + "\n").encode("utf-8")
+    payload_digest = hashlib.sha256(payload).hexdigest()
+    descriptor = None
+    session_descriptor = None
+    publication_descriptor = None
+    temporary = None
+    published = False
+    publication = "created"
+    result_digest = payload_digest
+    durability_warnings = []
+    temporary_cleanup_error = None
+    try:
+        session_flags = os.O_RDONLY
+        session_flags |= getattr(os, "O_DIRECTORY", 0)
+        session_flags |= getattr(os, "O_NOFOLLOW", 0)
+        session_path_identity = os.stat(session_path, follow_symlinks=False)
+        session_descriptor = os.open(session_path, session_flags)
+        session_fd_identity = os.fstat(session_descriptor)
+        if (
+            not stat.S_ISDIR(session_path_identity.st_mode)
+            or _component_identity(session_path_identity)
+            != _component_identity(session_fd_identity)
+        ):
+            return None, {
+                "code": "admission_publish_failed",
+                "message": "review session changed before ADMISSION.json publication",
+            }
+        descriptor, temporary = tempfile.mkstemp(prefix=".ADMISSION.", dir=session_path)
+        with os.fdopen(descriptor, "wb", closefd=True) as stream:
+            descriptor = None
+            stream.write(payload)
+            stream.flush()
+            os.fsync(stream.fileno())
+        publication_flags = os.O_RDONLY
+        publication_flags |= getattr(os, "O_NONBLOCK", 0)
+        publication_flags |= getattr(os, "O_NOFOLLOW", 0)
+        publication_descriptor = os.open(temporary, publication_flags)
+        try:
+            # A same-directory hard link is an atomic create-if-absent publication.
+            # Unlike replace/rename, it can never overwrite an admitted session.
+            os.link(temporary, target, follow_symlinks=False)
+            published = True
+            temporary_identity = _stat_identity(os.fstat(publication_descriptor))
+            repo_root, root_error = canonical_repo_root(session_path)
+            if root_error:
+                return None, {
+                    "code": "admission_publication_changed",
+                    "message": (
+                        "ADMISSION.json publication cannot be verified from the canonical "
+                        f"session path: {root_error}"
+                    ),
+                }
+            published_snapshot, published_error, _missing = stable_read_repo_bytes(
+                repo_root,
+                f".ultra/reviews/{session_path.resolve().name}/ADMISSION.json",
+                "published ADMISSION.json",
+            )
+            if (
+                published_error
+                or published_snapshot["identity"] != temporary_identity
+                or published_snapshot["bytes"] != payload
+            ):
+                detail = published_error or (
+                    "the canonical receipt identity or bytes differ from the newly "
+                    "published file"
+                )
+                return None, {
+                    "code": "admission_publication_changed",
+                    "message": (
+                        "ADMISSION.json canonical session or receipt changed during its "
+                        f"post-publication fresh rewalk: {detail}"
+                    ),
+                }
+        except FileExistsError:
+            repo_root, root_error = canonical_repo_root(session_path)
+            if root_error:
+                return None, {
+                    "code": "admission_conflict",
+                    "message": f"existing ADMISSION.json cannot be verified: {root_error}",
+                }
+            existing, existing_error, _missing = stable_read_repo_bytes(
+                repo_root,
+                f".ultra/reviews/{session_path.resolve().name}/ADMISSION.json",
+                "existing ADMISSION.json",
+            )
+            if existing_error or existing["bytes"] != payload:
+                detail = existing_error or "its bytes differ from the newly observed receipt"
+                return None, {
+                    "code": "admission_conflict",
+                    "message": (
+                        "ADMISSION.json is immutable once published and the existing "
+                        f"receipt cannot be reused because {detail}; start a fresh review session"
+                    ),
+                }
+            published = True
+            publication = "existing"
+            result_digest = existing["sha256"]
+        except OSError as error:
+            return None, {
+                "code": "admission_publish_failed",
+                "message": f"ADMISSION.json could not be published atomically: {error}",
+            }
+
+        removed, cleanup_error = cleanup_admission_temporary(
+            session_descriptor,
+            temporary,
+            publication_descriptor,
+        )
+        if removed:
+            temporary = None
+        else:
+            temporary_cleanup_error = cleanup_error
+
+        try:
+            os.fsync(session_descriptor)
+        except OSError as error:
+            durability_warnings.append({
+                "code": "admission_directory_fsync_failed",
+                "message": (
+                    "ADMISSION.json is published and usable, but directory durability "
+                    f"could not be confirmed: {error}"
+                ),
+            })
+    except OSError as error:
+        if published:
+            if not any(
+                warning["code"] == "admission_directory_fsync_failed"
+                for warning in durability_warnings
+            ):
+                durability_warnings.append({
+                    "code": "admission_directory_fsync_failed",
+                    "message": (
+                        "ADMISSION.json is published and usable, but directory durability "
+                        f"could not be confirmed: {error}"
+                    ),
+                })
+        else:
+            return None, {
+                "code": "admission_publish_failed",
+                "message": f"ADMISSION.json could not be published atomically: {error}",
+            }
+    finally:
+        if descriptor is not None:
+            os.close(descriptor)
+        if (
+            temporary is not None
+            and session_descriptor is not None
+            and publication_descriptor is not None
+        ):
+            removed, cleanup_error = cleanup_admission_temporary(
+                session_descriptor,
+                temporary,
+                publication_descriptor,
+            )
+            if removed:
+                temporary = None
+                temporary_cleanup_error = None
+            else:
+                temporary_cleanup_error = cleanup_error
+        if publication_descriptor is not None:
+            os.close(publication_descriptor)
+        if session_descriptor is not None:
+            os.close(session_descriptor)
+    if temporary is not None and temporary_cleanup_error is not None:
+        durability_warnings.append({
+            "code": "admission_temp_cleanup_failed",
+            "message": (
+                "ADMISSION.json is published and usable, but its temporary pathname "
+                f"could not be removed: {temporary_cleanup_error}"
+            ),
+        })
+    result = {
+        "digest": result_digest,
+        "publication": publication,
+    }
+    if len(durability_warnings) == 1:
+        result["durability_warning"] = durability_warnings[0]
+    elif durability_warnings:
+        result["durability_warning"] = {
+            "code": "admission_multiple_durability_warnings",
+            "message": (
+                "ADMISSION.json is published and usable, but multiple durability "
+                "operations remain unresolved"
+            ),
+            "warnings": durability_warnings,
+        }
+    return result, None
+
+
+def validate_admission_receipt(receipt, packet):
+    if not isinstance(receipt, dict) or set(receipt) != ADMISSION_FIELDS:
+        return "ADMISSION.json must contain exactly the strict admission receipt fields"
+    contract = packet["data"].get("admission_contract")
+    if contract == ADMISSION_CONTRACT_V1:
+        expected_schema = ADMISSION_SCHEMA_V1
+        expected_version = 1
+        observation_fields = ADMISSION_OBSERVATION_FIELDS_V1
+    elif contract == ADMISSION_CONTRACT_V2:
+        expected_schema = ADMISSION_SCHEMA_V2
+        expected_version = 2
+        observation_fields = ADMISSION_OBSERVATION_FIELDS_V2
+    else:
+        return "ADMISSION.json requires a recognized strict packet admission_contract"
+    if (
+        receipt.get("$schema") != expected_schema
+        or receipt.get("version") != expected_version
+    ):
+        return (
+            f"ADMISSION.json must use {expected_schema} version {expected_version} "
+            "for its packet admission_contract"
+        )
+    if receipt.get("session") != packet["data"]["session"]:
+        return "ADMISSION.json session does not match WORKER-PACKET.json"
+    if receipt.get("packet_digest") != packet["digest"]:
+        return "ADMISSION.json packet_digest does not match WORKER-PACKET.json"
+    if receipt.get("head") != packet["data"]["head"]:
+        return "ADMISSION.json head does not match WORKER-PACKET.json"
+    if not digest_string(receipt.get("subject_digest")):
+        return "ADMISSION.json subject_digest must be a lowercase SHA-256 digest"
+
+    observations = receipt.get("observations")
+    if not isinstance(observations, list):
+        return "ADMISSION.json observations must be an array"
+    for index, observation in enumerate(observations):
+        if not isinstance(observation, dict) or set(observation) != observation_fields:
+            return f"ADMISSION.json observations[{index}] has invalid fields"
+        if observation.get("role") not in {
+            "context", "north_star", "change", "acceptance_source",
+        }:
+            return f"ADMISSION.json observations[{index}] has an invalid role"
+        _, path_error = validate_repo_path(
+            observation.get("path"),
+            packet["repo_root"],
+            f"ADMISSION.json observations[{index}].path",
+        )
+        if path_error:
+            return path_error
+        if not digest_string(observation.get("sha256")):
+            return f"ADMISSION.json observations[{index}].sha256 is invalid"
+        if expected_version == 2:
+            byte_length = observation.get("byte_length")
+            if (
+                not isinstance(byte_length, int)
+                or isinstance(byte_length, bool)
+                or byte_length < 0
+            ):
+                return f"ADMISSION.json observations[{index}].byte_length is invalid"
+
+    subject_projection = (
+        packet["data"]["subject_observations"][:-2]
+        if expected_version == 2
+        else [
+            {
+                "role": "change",
+                "path": f".ultra/changes/active/{packet['data']['change_id']}/intent.md",
+            },
+            *(
+                {"role": "acceptance_source", "path": path}
+                for path in acceptance_source_paths(packet["data"])
+            ),
+        ]
+    )
+    expected_roles_and_paths = [
+        *(('context', item["path"]) for item in packet["data"]["context_files"]),
+        ("north_star", ".ultra/north-star.md"),
+        *((item["role"], item["path"]) for item in subject_projection),
+    ]
+    observed_roles_and_paths = [
+        (item["role"], item["path"]) for item in observations
+    ]
+    if observed_roles_and_paths != expected_roles_and_paths:
+        return "ADMISSION.json observations are not the exact packet-derived subject projection"
+    for index, context_file in enumerate(packet["data"]["context_files"]):
+        if observations[index]["sha256"] != context_file["sha256"]:
+            return "ADMISSION.json context observation does not match WORKER-PACKET.json"
+    if expected_version == 2:
+        subject_offset = len(packet["data"]["context_files"]) + 1
+        if observations[subject_offset:] != subject_projection:
+            return (
+                "ADMISSION.json Change and acceptance observations do not exactly "
+                "match WORKER-PACKET.json subject_observations"
+            )
+
+    report_receipt = receipt.get("north_star_report")
+    if (
+        not isinstance(report_receipt, dict)
+        or set(report_receipt) != NORTH_STAR_REPORT_RECEIPT_FIELDS
+        or report_receipt.get("schema") != "ultra-north-star-validation-v1"
+        or not digest_string(report_receipt.get("sha256"))
+        or not digest_string(report_receipt.get("input_sha256"))
+        or not isinstance(report_receipt.get("input_byte_length"), int)
+        or isinstance(report_receipt.get("input_byte_length"), bool)
+        or report_receipt["input_byte_length"] < 0
+    ):
+        return "ADMISSION.json north_star_report receipt is invalid"
+    source_observations = report_receipt.get("source_observations")
+    if not isinstance(source_observations, list) or len(source_observations) != 2:
+        return "ADMISSION.json north_star_report source observations are invalid"
+    observed_source_roles = []
+    for index, observation in enumerate(source_observations):
+        if (
+            not isinstance(observation, dict)
+            or set(observation) != NORTH_STAR_SOURCE_OBSERVATION_FIELDS
+            or observation.get("role") not in {"decision", "snapshot"}
+            or not digest_string(observation.get("sha256"))
+            or not isinstance(observation.get("byte_length"), int)
+            or isinstance(observation.get("byte_length"), bool)
+            or observation["byte_length"] < 0
+        ):
+            return (
+                "ADMISSION.json north_star_report "
+                f"source_observations[{index}] is invalid"
+            )
+        _, path_error = validate_repo_path(
+            observation.get("path"),
+            packet["repo_root"],
+            f"ADMISSION.json north_star_report source_observations[{index}].path",
+        )
+        if path_error:
+            return path_error
+        observed_source_roles.append(observation["role"])
+    if observed_source_roles != ["decision", "snapshot"]:
+        return (
+            "ADMISSION.json north_star_report source observations must be the exact "
+            "decision then snapshot pair"
+        )
+    if (
+        expected_version == 2
+        and source_observations != packet["data"]["subject_observations"][-2:]
+    ):
+        return (
+            "ADMISSION.json canonical report decision and snapshot observations do not "
+            "exactly match WORKER-PACKET.json subject_observations"
+        )
+    north_star_observation = next(
+        item for item in observations if item["role"] == "north_star"
+    )
+    if report_receipt["input_sha256"] != north_star_observation["sha256"]:
+        return "ADMISSION.json canonical report input does not match its North Star observation"
+    expected_subject_digest = canonical_json_digest(canonical_admission_subject(receipt))
+    if receipt["subject_digest"] != expected_subject_digest:
+        return "ADMISSION.json subject_digest does not match its exact admission observations"
+    return None
+
+
+def load_admission_receipt(session_path: Path, packet):
+    admission_repo_path = (
+        f".ultra/reviews/{packet['data']['session']}/ADMISSION.json"
+    )
+    receipt, snapshot, error, _missing = stable_read_repo_json(
+        packet["repo_root"],
+        admission_repo_path,
+        "ADMISSION.json",
+    )
+    if error:
+        return None, error
+    error = validate_admission_receipt(receipt, packet)
+    if error:
+        return None, error
+    return {"data": receipt, "digest": snapshot["sha256"]}, None
+
+
+def _foreign_review_references(value, own_session: str):
+    """Collect every foreign review-session path carried anywhere in a value."""
+    found = []
+    if isinstance(value, str):
+        for match in FOREIGN_REVIEW_SESSION.finditer(value):
+            if match.group(1) != own_session:
+                found.append(match.group(0))
+    elif isinstance(value, dict):
+        for key, item in value.items():
+            found.extend(_foreign_review_references(key, own_session))
+            found.extend(_foreign_review_references(item, own_session))
+    elif isinstance(value, list):
+        for item in value:
+            found.extend(_foreign_review_references(item, own_session))
+    return found
+
+
+def validate_packet_review_history(packet, repo_root: Path) -> str | None:
+    """Bound packet review history to one fully validated direct parent.
+
+    The only foreign review path allowed anywhere in a task delta packet is the
+    exact `review_history.parent_summary_ref` field; an initial packet carries
+    no foreign review reference at all. Admission then observes the parent
+    SUMMARY as a bounded stable repository regular file, requires its exact
+    recorded digest, requires the parent to bind the same mode, change, and
+    tasks as the current packet, and accepts only still-unresolved current
+    P0/P1 blocker ids that exist in that parent. Transitive summary chains and
+    full-history replays stay out of new packets; the archive under
+    `.ultra/reviews/**` is read only for a dedicated incident or owner audit.
+    """
+    own_session = packet.get("session")
+    history = packet.get("review_history")
+    if history is not None and (
+        not isinstance(history, dict) or set(history) != REVIEW_HISTORY_FIELDS
+    ):
+        return (
+            "WORKER-PACKET.json review_history must be exactly one direct parent "
+            "{parent_summary_ref, parent_summary_digest, unresolved_blocking_ids}"
+        )
+    references: list[str] = []
+    if isinstance(packet, dict):
+        for key, item in packet.items():
+            if key == "review_history":
+                continue
+            references.extend(_foreign_review_references(key, own_session))
+            references.extend(_foreign_review_references(item, own_session))
+    if history is not None:
+        reference = history.get("parent_summary_ref")
+        if (
+            not isinstance(reference, str)
+            or not re.fullmatch(
+                r"\.ultra/reviews/[A-Za-z0-9][A-Za-z0-9._-]*/SUMMARY\.json",
+                reference,
+            )
+            or reference.split("/")[2] == own_session
+        ):
+            return (
+                "WORKER-PACKET.json review_history.parent_summary_ref must name "
+                "exactly one prior session's SUMMARY.json"
+            )
+        if not digest_string(history.get("parent_summary_digest")):
+            return (
+                "WORKER-PACKET.json review_history.parent_summary_digest must be a "
+                "lowercase SHA-256 digest"
+            )
+        blocking_ids = history.get("unresolved_blocking_ids")
+        if (
+            not isinstance(blocking_ids, list)
+            or not all(nonempty_string(item) for item in blocking_ids)
+        ):
+            return (
+                "WORKER-PACKET.json review_history.unresolved_blocking_ids must be "
+                "an array of finding-id strings"
+            )
+    if references:
+        return (
+            "WORKER-PACKET.json may reference a foreign review session only through "
+            "review_history.parent_summary_ref; observed outside that field: "
+            + ", ".join(sorted(set(references)))
+        )
+    if history is None:
+        return None
+
+    parent_snapshot, parent_error, _missing = stable_read_repo_bytes(
+        repo_root,
+        history["parent_summary_ref"],
+        "review_history parent SUMMARY",
+    )
+    if parent_error:
+        return (
+            "WORKER-PACKET.json review_history parent summary could not be observed "
+            f"as a stable repository regular file: {parent_error}"
+        )
+    if parent_snapshot["sha256"] != history["parent_summary_digest"]:
+        return (
+            "WORKER-PACKET.json review_history.parent_summary_digest does not match "
+            "the observed parent SUMMARY bytes"
+        )
+    try:
+        parent = json.loads(parent_snapshot["bytes"].decode("utf-8"))
+    except (UnicodeDecodeError, json.JSONDecodeError) as error:
+        return (
+            "WORKER-PACKET.json review_history parent SUMMARY is invalid UTF-8 "
+            f"JSON: {error}"
+        )
+    if not isinstance(parent, dict) or parent.get("$schema") != "ultra-review-summary-v4":
+        return (
+            "WORKER-PACKET.json review_history parent SUMMARY must be an "
+            "ultra-review-summary-v4 object"
+        )
+    for field, current in (
+        ("mode", packet.get("mode")),
+        ("change_id", packet.get("change_id")),
+        ("task_ids", packet.get("task_ids")),
+    ):
+        if parent.get(field) != current:
+            return (
+                "WORKER-PACKET.json review_history parent SUMMARY "
+                f"{field} does not match the current packet"
+            )
+    findings = parent.get("findings")
+    if not isinstance(findings, list):
+        return (
+            "WORKER-PACKET.json review_history parent SUMMARY carries no findings "
+            "array to resolve blocking ids against"
+        )
+    severity_by_id: dict[str, str] = {}
+    for finding in findings:
+        if isinstance(finding, dict) and nonempty_string(finding.get("id")):
+            severity_by_id[finding["id"]] = finding.get("severity")
+    blocking_ids = history["unresolved_blocking_ids"]
+    if len(blocking_ids) != len(set(blocking_ids)):
+        return (
+            "WORKER-PACKET.json review_history.unresolved_blocking_ids must not "
+            "repeat a finding id"
+        )
+    for blocking_id in blocking_ids:
+        severity = severity_by_id.get(blocking_id)
+        if severity is None:
+            return (
+                "WORKER-PACKET.json review_history.unresolved_blocking_ids names "
+                f"`{blocking_id}`, which is not a finding of the parent SUMMARY"
+            )
+        if severity not in {"P0", "P1"}:
+            return (
+                "WORKER-PACKET.json review_history.unresolved_blocking_ids names "
+                f"`{blocking_id}`, whose parent severity `{severity}` is not a "
+                "current P0/P1 blocker"
+            )
+    return None
+
+
+def load_worker_packet(
+    session_path: Path,
+    expected_packet_digest: str,
+    admit_subjects=False,
+    require_admission=False,
+    legacy_v4=False,
+):
+    repo_root, root_error = canonical_repo_root(session_path)
+    if root_error:
+        return None, f"WORKER-PACKET.json {root_error}"
+    session_name = session_path.resolve().name
+    packet_repo_path = f".ultra/reviews/{session_name}/WORKER-PACKET.json"
+    packet_snapshot, packet_error, _missing = stable_read_repo_bytes(
+        repo_root,
+        packet_repo_path,
+        "WORKER-PACKET.json",
+    )
+    if packet_error:
+        return None, packet_error
+    packet_bytes = packet_snapshot["bytes"]
+    actual_digest = packet_snapshot["sha256"]
+    if actual_digest != expected_packet_digest:
+        return None, (
+            "WORKER-PACKET.json bytes do not match expected immutable packet digest "
+            f"{expected_packet_digest}; observed {actual_digest}"
+        )
+    try:
+        packet = json.loads(packet_bytes.decode("utf-8"))
+    except (UnicodeDecodeError, json.JSONDecodeError) as error:
+        return None, f"WORKER-PACKET.json is invalid UTF-8 JSON: {error}"
+    if not isinstance(packet, dict):
+        return None, "WORKER-PACKET.json top-level value must be an object"
+    missing_fields = sorted(PACKET_REQUIRED_FIELDS - set(packet))
+    if missing_fields:
+        return None, (
+            "WORKER-PACKET.json missing required fields: " + ", ".join(missing_fields)
+        )
+    if packet.get("$schema") != "ultra-review-worker-packet-v1":
+        return None, "WORKER-PACKET.json must use ultra-review-worker-packet-v1"
+    admission_contract = packet.get("admission_contract")
+    if admit_subjects:
+        if admission_contract != ADMISSION_CONTRACT_V2:
+            return None, (
+                "WORKER-PACKET.json admission_contract must be "
+                f"{ADMISSION_CONTRACT_V2} for new strict v4 admission"
+            )
+    elif require_admission and admission_contract not in STRICT_ADMISSION_CONTRACTS:
+        return None, (
+            "WORKER-PACKET.json admission_contract must declare retained strict v1 "
+            "or current strict v2 for strict v4 consumption"
+        )
+    elif legacy_v4 and admission_contract in STRICT_ADMISSION_CONTRACTS:
+        return None, (
+            "WORKER-PACKET.json declares a strict admission_contract and "
+            "cannot be consumed with --legacy-v4"
+        )
+    for field in ["session", "head", "range", "change_id", "output_directory"]:
+        if not nonempty_string(packet.get(field)):
+            return None, f"WORKER-PACKET.json {field} must be a non-empty string"
+    if (
+        not ARTIFACT_STEM.fullmatch(packet["session"])
+        or packet["session"] != session_path.resolve().name
+    ):
+        return None, (
+            "WORKER-PACKET.json session must be a safe stem matching the canonical "
+            "session directory"
+        )
+    if not timestamp_string(packet.get("created_at")):
+        return None, "WORKER-PACKET.json created_at must be an RFC 3339 timestamp"
+    if not git_blob_digest_string(packet.get("head")):
+        return None, "WORKER-PACKET.json head must be a lowercase 40-hex Git commit digest"
+    if packet.get("mode") not in REVIEW_MODES:
+        return None, "WORKER-PACKET.json mode must be task, change, or plan"
+    for field, require_nonempty in [
+        ("task_ids", packet["mode"] == "task"),
+        ("acceptance", True),
+        ("public_seams", True),
+    ]:
+        array_error = validate_string_array(
+            packet.get(field),
+            f"WORKER-PACKET.json {field}",
+            require_nonempty,
+        )
+        if array_error:
+            return None, array_error
+    for index, acceptance in enumerate(packet["acceptance"]):
+        source, separator, literal = acceptance.partition(": ")
+        if not separator or not literal:
+            return None, (
+                f"WORKER-PACKET.json acceptance[{index}] must be "
+                "<repository-path>#<heading>: <captured owner-readable claim>"
+            )
+        _, path_error = validate_repo_path(
+            source,
+            repo_root,
+            f"WORKER-PACKET.json acceptance[{index}] source",
+            allow_fragment=True,
+        )
+        if path_error:
+            return None, path_error
+
+    output_directory, path_error = validate_repo_path(
+        packet["output_directory"],
+        repo_root,
+        "WORKER-PACKET.json output_directory",
+    )
+    if path_error:
+        return None, path_error
+    canonical_output = f".ultra/reviews/{packet['session']}"
+    if output_directory != canonical_output:
+        return None, (
+            "WORKER-PACKET.json output_directory must equal the canonical session path "
+            f"{canonical_output}"
+        )
+
+    trace_data = packet.get("north_star_trace")
+    if not isinstance(trace_data, dict) or set(trace_data) != TRACE_FIELDS:
+        return None, (
+            "WORKER-PACKET.json north_star_trace must contain exactly path, "
+            "first_principles, serves, touches, north_star_revision, and "
+            "north_star_digest"
+        )
+    trace = _trace_arrays(packet.get("north_star_trace"))
+    if trace is None:
+        return None, "WORKER-PACKET.json north_star_trace arrays are invalid"
+    _, path_error = validate_repo_path(
+        trace_data.get("path"),
+        repo_root,
+        "WORKER-PACKET.json north_star_trace.path",
+        allow_fragment=True,
+    )
+    if path_error:
+        return None, path_error
+    trace_source, fragment = trace_data["path"].split("#", 1)
+    parts = trace_source.split("/")
+    if (
+        fragment != "north-star-trace"
+        or len(parts) != 5
+        or parts[:2] != [".ultra", "changes"]
+        or parts[2] not in {"active", "archive", "abandoned"}
+        or parts[3] != packet["change_id"]
+        or parts[4] != "intent.md"
+    ):
+        return None, (
+            "north_star_trace.path must bind change_id to "
+            ".ultra/changes/<active|archive|abandoned>/<change_id>/intent.md"
+            "#north-star-trace"
+        )
+    if not nonempty_string(trace_data.get("north_star_revision")):
+        return None, (
+            "WORKER-PACKET.json north_star_trace.north_star_revision must be a "
+            "non-empty string"
+        )
+    if not git_blob_digest_string(trace_data.get("north_star_digest")):
+        return None, (
+            "WORKER-PACKET.json north_star_trace.north_star_digest must be a "
+            "lowercase 40-hex Git blob digest"
+        )
+    if admission_contract == ADMISSION_CONTRACT_V2:
+        subject_observation_error = validate_packet_subject_observations(packet, repo_root)
+        if subject_observation_error:
+            return None, subject_observation_error
+
+    context_files = packet.get("context_files")
+    if not isinstance(context_files, list):
+        return None, "WORKER-PACKET.json context_files must be an array"
+    seen_context_paths = set()
+    for index, context_file in enumerate(context_files):
+        if not isinstance(context_file, dict) or set(context_file) != CONTEXT_FIELDS:
+            return None, (
+                f"WORKER-PACKET.json context_files[{index}] must contain exactly "
+                "path and sha256"
+            )
+        context_path, path_error = validate_repo_path(
+            context_file.get("path"),
+            repo_root,
+            f"WORKER-PACKET.json context_files[{index}].path",
+        )
+        if path_error:
+            return None, path_error
+        if context_path in seen_context_paths:
+            return None, "WORKER-PACKET.json context_files paths must be unique"
+        seen_context_paths.add(context_path)
+        if not digest_string(context_file.get("sha256")):
+            return None, (
+                f"WORKER-PACKET.json context_files[{index}].sha256 must be a "
+                "lowercase SHA-256 digest"
+            )
+    if packet["mode"] == "task":
+        expected_context_paths = [
+            f".ultra/contexts/task-{task_id}.md" for task_id in packet["task_ids"]
+        ]
+        observed_context_paths = [item["path"] for item in context_files]
+        if observed_context_paths != expected_context_paths:
+            return None, (
+                "task-mode context_files paths must exactly follow ordered task_ids: "
+                + ", ".join(expected_context_paths)
+            )
+
+    diff_files = packet.get("diff_files")
+    if not isinstance(diff_files, list) or not diff_files:
+        return None, "WORKER-PACKET.json diff_files must be a non-empty path array"
+    normalized_diff_files = []
+    for index, diff_file in enumerate(diff_files):
+        normalized_path, path_error = validate_repo_path(
+            diff_file,
+            repo_root,
+            f"WORKER-PACKET.json diff_files[{index}]",
+        )
+        if path_error:
+            return None, path_error
+        normalized_diff_files.append(normalized_path)
+    if len(normalized_diff_files) != len(set(normalized_diff_files)):
+        return None, "WORKER-PACKET.json diff_files must not contain duplicates"
+
+    history_error = validate_packet_review_history(packet, repo_root)
+    if history_error:
+        return None, history_error
+
+    workers = packet.get("workers")
+    if not isinstance(workers, list) or not workers:
+        return None, "WORKER-PACKET.json workers must be a non-empty array"
+    by_stem = {}
+    by_agent = {}
+    ordered = []
+    previous_order = -1
+    for index, worker in enumerate(workers):
+        if not isinstance(worker, dict) or set(worker) != WORKER_FIELDS:
+            return None, (
+                f"WORKER-PACKET.json workers[{index}] must contain exactly agent, "
+                "axis, lens, and output"
+            )
+        agent = worker.get("agent")
+        axis = worker.get("axis")
+        lens = worker.get("lens")
+        output = worker.get("output")
+        if agent not in REVIEW_WORKERS or agent in by_agent:
+            return None, f"WORKER-PACKET.json has invalid or duplicate worker agent: {agent}"
+        canonical = CANONICAL_WORKER_BY_AGENT[agent]
+        if axis != canonical["axis"]:
+            return None, (
+                f"WORKER-PACKET.json worker {agent} axis must use canonical value "
+                f"{canonical['axis']}"
+            )
+        if lens != canonical["lens"]:
+            return None, (
+                f"WORKER-PACKET.json worker {agent} lens must use canonical path "
+                f"{canonical['lens']}"
+            )
+        _, path_error = validate_repo_path(
+            lens,
+            repo_root,
+            f"WORKER-PACKET.json worker {agent} lens",
+        )
+        if path_error:
+            return None, path_error
+        output_path, path_error = validate_repo_path(
+            output,
+            repo_root,
+            f"WORKER-PACKET.json worker {agent} output",
+        )
+        if path_error:
+            return None, path_error
+        canonical_worker_output = f"{output_directory}/{agent}.json"
+        if output_path != canonical_worker_output:
+            return None, (
+                f"WORKER-PACKET.json worker {agent} output must equal canonical path "
+                f"{canonical_worker_output}"
+            )
+        if canonical["order"] <= previous_order:
+            return None, "WORKER-PACKET.json workers must use canonical roster order"
+        previous_order = canonical["order"]
+        output_name = PurePosixPath(output_path).name
+        output_stem = PurePosixPath(output_path).stem
+        if output_stem in by_stem:
+            return None, f"WORKER-PACKET.json repeats output stem: {output_stem}"
+        normalized = {
+            "agent": agent,
+            "axis": axis,
+            "lens": lens,
+            "output": output_path,
+            "stem": output_stem,
+            "filename": output_name,
+        }
+        by_stem[output_stem] = normalized
+        by_agent[agent] = normalized
+        ordered.append(normalized)
+    if "review-spec" not in by_agent:
+        return None, (
+            "WORKER-PACKET.json workers must include canonical review-spec for the "
+            "permanent specification axis"
+        )
+    admission = None
+    if admit_subjects:
+        admission, subject_error = validate_packet_admission(
+            packet,
+            repo_root,
+            actual_digest,
+        )
+        if subject_error:
+            return None, f"WORKER-PACKET.json subject validation failed: {subject_error}"
+    loaded = {
+        "data": packet,
+        "digest": actual_digest,
+        "trace": {field: set(values) for field, values in trace.items()},
+        "repo_root": repo_root,
+        "diff_scope": set(normalized_diff_files),
+        "context_digest": canonical_context_digest(context_files),
+        "admission": admission,
+        "workers": ordered,
+        "by_stem": by_stem,
+        "by_agent": by_agent,
+    }
+    if require_admission:
+        receipt, receipt_error = load_admission_receipt(session_path, loaded)
+        if receipt_error:
+            return None, (
+                "strict v4 packet consumption requires a valid ADMISSION.json receipt "
+                f"from packet mode: {receipt_error}"
+            )
+        loaded["admission_receipt"] = receipt
+    return loaded, None
+
+
+def reload_worker_packet(
+    session_path: Path,
+    expected_packet_digest: str,
+    admit_subjects=False,
+    require_admission=False,
+    legacy_v4=False,
+):
+    return load_worker_packet(
+        session_path,
+        expected_packet_digest,
+        admit_subjects,
+        require_admission,
+        legacy_v4,
+    )
+
+
+def packet_admission_binding(packet):
+    receipt = packet.get("admission_receipt") if packet is not None else None
+    if receipt is None:
+        return None
+    return {
+        "admission_digest": receipt["digest"],
+        "subject_digest": receipt["data"]["subject_digest"],
+    }
+
+
+def validate_artifact_admission_binding(data, packet, label: str):
+    expected = packet_admission_binding(packet)
+    if expected is None:
+        return None
+    for field in ("admission_digest", "subject_digest"):
+        if not digest_string(data.get(field)):
+            return f"{label} {field} must be the lowercase SHA-256 from ADMISSION.json"
+        if data[field] != expected[field]:
+            return f"{label} {field} must exactly match ADMISSION.json"
+    return None
+
+
+def validate_pinned_admission(packet, expected):
+    if expected is None:
+        return None
+    if packet_admission_binding(packet) != expected:
+        return (
+            "ADMISSION.json receipt changed during polling; the strict review session "
+            "is INCOMPLETE and must be restarted as a fresh session"
+        )
+    return None
+
+
+def validate_north_star_trace(value, allowed=None):
+    if not isinstance(value, dict) or set(value) != {"first_principles", "serves", "touches"}:
+        return "finding north_star_trace must contain exactly first_principles, serves, and touches"
+    patterns = {
+        "first_principles": r"FP-[A-Za-z0-9][A-Za-z0-9._-]*",
+        "serves": r"NS-[A-Za-z0-9][A-Za-z0-9._-]*",
+        "touches": r"HC-[A-Za-z0-9][A-Za-z0-9._-]*",
+    }
+    for field, pattern in patterns.items():
+        values = value.get(field)
+        if not isinstance(values, list) or not all(
+            isinstance(item, str) and re.fullmatch(pattern, item) for item in values
+        ):
+            return f"finding north_star_trace.{field} must be an array of resolving {pattern[:2]} IDs"
+        if len(values) != len(set(values)):
+            return f"finding north_star_trace.{field} must not contain duplicates"
+        if allowed is not None:
+            outside = [item for item in values if item not in allowed[field]]
+            if outside:
+                return (
+                    f"finding north_star_trace.{field} contains IDs outside the immutable "
+                    f"Worker Packet: {', '.join(outside)}"
+                )
+    return None
+
+
+def validate_finding(
+    finding,
+    artifact_axis: str,
+    seen_ids,
+    findings_schema: str,
+    allowed_trace=None,
+    allowed_files=None,
+    repo_root=None,
+):
     if not isinstance(finding, dict):
         return "finding must be an object"
     missing = sorted(FINDING_FIELDS - set(finding))
@@ -88,40 +1999,107 @@ def validate_finding(finding, artifact_axis: str, seen_ids):
     for field in ["category", "title", "file", "trigger", "impact", "evidence", "suggestion"]:
         if not nonempty_string(finding.get(field)):
             return f"finding {field} must be a non-empty string"
+    if repo_root is not None:
+        finding_file, path_error = validate_repo_path(
+            finding["file"],
+            repo_root,
+            "finding file",
+        )
+        if path_error:
+            return path_error
+        if finding_file not in allowed_files:
+            return "finding file must be within the immutable Worker Packet diff scope"
     if not isinstance(finding.get("line"), int) or finding["line"] < 1:
         return "finding line must be a positive integer"
     line_end = finding.get("line_end")
     if line_end is not None and (not isinstance(line_end, int) or line_end < finding["line"]):
         return "finding line_end must be at or after line"
+    if findings_schema == FINDINGS_SCHEMA:
+        if "north_star_trace" not in finding:
+            return "finding missing fields: north_star_trace"
+        trace_error = validate_north_star_trace(
+            finding.get("north_star_trace"),
+            allowed_trace,
+        )
+        if trace_error:
+            return trace_error
     return None
 
 
-def validate_specialist(data, expected_packet_digest: str):
-    if data.get("$schema") != FINDINGS_SCHEMA:
-        return f"$schema must be {FINDINGS_SCHEMA}"
+def validate_specialist(
+    data,
+    expected_packet_digest: str,
+    findings_schema: str,
+    expected_worker=None,
+    packet=None,
+):
+    if data.get("$schema") != findings_schema:
+        return f"$schema must be {findings_schema}"
     if not nonempty_string(data.get("agent")):
         return "agent must be a non-empty string"
+    if expected_worker is not None and data.get("agent") != expected_worker["agent"]:
+        return f"agent must match packet roster worker {expected_worker['agent']}"
     if not digest_string(data.get("packet_digest")):
         return "packet_digest must be a lowercase SHA-256 digest"
     if data["packet_digest"] != expected_packet_digest:
         return f"packet_digest does not match expected immutable packet {expected_packet_digest}"
+    admission_error = validate_artifact_admission_binding(
+        data,
+        packet,
+        "specialist artifact",
+    )
+    if admission_error:
+        return admission_error
     axis = data.get("axis")
     if axis not in AXES:
         return "axis must be spec_fidelity or engineering_standards"
-    for field in ["session", "timestamp"]:
-        if not nonempty_string(data.get(field)):
-            return f"{field} must be a non-empty string"
+    if expected_worker is not None and axis != expected_worker["axis"]:
+        return f"axis must match packet roster axis {expected_worker['axis']}"
+    if not nonempty_string(data.get("session")):
+        return "session must be a non-empty string"
+    if findings_schema == FINDINGS_SCHEMA:
+        if not timestamp_string(data.get("timestamp")):
+            return "timestamp must be an RFC 3339 timestamp"
+    elif not nonempty_string(data.get("timestamp")):
+        return "timestamp must be a non-empty string"
+    if packet is not None and data.get("session") != packet["data"]["session"]:
+        return "session must match WORKER-PACKET.json"
     scope = data.get("scope")
     if not isinstance(scope, dict):
         return "scope must be an object"
     for field in ["head", "range"]:
         if not nonempty_string(scope.get(field)):
             return f"scope.{field} must be a non-empty string"
+        if packet is not None and scope.get(field) != packet["data"][field]:
+            return f"scope.{field} must match WORKER-PACKET.json"
     if not isinstance(scope.get("files_analyzed"), list) or not all(
         nonempty_string(item) for item in scope["files_analyzed"]
     ):
         return "scope.files_analyzed must be a string array"
-    if not isinstance(scope.get("diff_only"), bool):
+    if packet is not None:
+        normalized_analyzed = []
+        for index, analyzed_file in enumerate(scope["files_analyzed"]):
+            normalized_path, path_error = validate_repo_path(
+                analyzed_file,
+                packet["repo_root"],
+                f"scope.files_analyzed[{index}]",
+            )
+            if path_error:
+                return path_error
+            normalized_analyzed.append(normalized_path)
+        if len(normalized_analyzed) != len(set(normalized_analyzed)):
+            return "scope.files_analyzed must not contain duplicates"
+        outside_scope = [
+            item for item in normalized_analyzed if item not in packet["diff_scope"]
+        ]
+        if outside_scope:
+            return (
+                "scope.files_analyzed must stay within the immutable Worker Packet "
+                "diff scope: " + ", ".join(outside_scope)
+            )
+        if scope.get("diff_only") is not True:
+            return "scope.diff_only must be true for packet-bound v4 artifacts"
+    elif not isinstance(scope.get("diff_only"), bool):
         return "scope.diff_only must be boolean"
     if data.get("status") != "complete":
         return "status must be complete"
@@ -130,51 +2108,197 @@ def validate_specialist(data, expected_packet_digest: str):
         return "findings must be an array"
     seen_ids = set()
     for finding in findings:
-        error = validate_finding(finding, axis, seen_ids)
+        error = validate_finding(
+            finding,
+            axis,
+            seen_ids,
+            findings_schema,
+            packet["trace"] if packet is not None else None,
+            packet["diff_scope"] if packet is not None else None,
+            packet["repo_root"] if packet is not None else None,
+        )
         if error:
             return error
     for field in ["positive_observations", "limitations"]:
         if not isinstance(data.get(field), list):
             return f"{field} must be an array"
+    coverage_refs = data.get("coverage_refs")
+    if not isinstance(coverage_refs, list) or not all(
+        nonempty_string(item) for item in coverage_refs
+    ):
+        return "coverage_refs must be a string array"
+    if not findings and not coverage_refs and not data["limitations"]:
+        return "zero findings require coverage_refs or an explicit limitation"
     return None
 
 
-def evaluate_artifacts(session_path: Path, expected, expected_packet_digest: str):
+def evaluate_artifacts(
+    session_path: Path,
+    expected,
+    expected_packet_digest: str,
+    findings_schema: str,
+    packet=None,
+):
     done = []
     missing = []
     invalid = []
     errors = {}
+    digests = {}
+    if packet is None:
+        repo_root, root_error = canonical_repo_root(session_path)
+        if root_error:
+            return [], [], list(expected), {"session": root_error}, {}
+    else:
+        repo_root = packet["repo_root"]
     for stem in expected:
-        path = session_path / f"{stem}.json"
-        if not path.exists():
+        expected_worker = packet["by_stem"].get(stem) if packet is not None else None
+        if packet is not None and expected_worker is None:
+            invalid.append(stem)
+            errors[stem] = "requested artifact stem is absent from WORKER-PACKET.json roster"
+            continue
+        repo_path = (
+            expected_worker["output"]
+            if expected_worker is not None
+            else f".ultra/reviews/{session_path.resolve().name}/{stem}.json"
+        )
+        data, snapshot, error, is_missing = stable_read_repo_json(
+            repo_root,
+            repo_path,
+            f"specialist artifact {stem}",
+        )
+        if is_missing:
             missing.append(stem)
             continue
-        data, error = read_json(path)
         if error is None:
-            error = validate_specialist(data, expected_packet_digest)
+            error = validate_specialist(
+                data,
+                expected_packet_digest,
+                findings_schema,
+                expected_worker,
+                packet,
+            )
         if error:
             invalid.append(stem)
             errors[stem] = error
         else:
             done.append(stem)
-    return done, missing, invalid, errors
+            digests[stem] = snapshot["sha256"]
+    return done, missing, invalid, errors, digests
 
 
-def wait_for_agents(session_path: Path, expected, expected_packet_digest: str, timeout: float) -> bool:
+def wait_for_agents(
+    session_path: Path,
+    expected,
+    expected_packet_digest: str,
+    findings_schema: str,
+    timeout: float,
+    require_admission: bool,
+    legacy_v4: bool = False,
+) -> bool:
+    packet = None
+    if findings_schema == FINDINGS_SCHEMA:
+        packet, packet_error = reload_worker_packet(
+            session_path,
+            expected_packet_digest,
+            require_admission=require_admission,
+            legacy_v4=legacy_v4,
+        )
+        if packet_error:
+            print(json.dumps({
+                "status": "incomplete",
+                "artifacts_done": [],
+                "artifacts_missing": [],
+                "artifacts_invalid": list(expected),
+                "errors": {"WORKER-PACKET": packet_error},
+                "count": 0,
+            }, sort_keys=True))
+            return False
+    pinned_admission = packet_admission_binding(packet)
     deadline = time.monotonic() + timeout
     while True:
-        done, missing, invalid, errors = evaluate_artifacts(
-            session_path, expected, expected_packet_digest
+        current_packet = packet
+        if packet is not None:
+            current_packet, packet_error = reload_worker_packet(
+                session_path,
+                expected_packet_digest,
+                require_admission=require_admission,
+                legacy_v4=legacy_v4,
+            )
+            if packet_error:
+                print(json.dumps({
+                    "status": "incomplete",
+                    "artifacts_done": [],
+                    "artifacts_missing": [],
+                    "artifacts_invalid": list(expected),
+                    "errors": {"WORKER-PACKET": packet_error},
+                    "count": 0,
+                }, sort_keys=True))
+                return False
+            admission_error = validate_pinned_admission(
+                current_packet,
+                pinned_admission,
+            )
+            if admission_error:
+                print(json.dumps({
+                    "status": "incomplete",
+                    "artifacts_done": [],
+                    "artifacts_missing": [],
+                    "artifacts_invalid": list(expected),
+                    "errors": {"ADMISSION": admission_error},
+                    "count": 0,
+                }, sort_keys=True))
+                return False
+        done, missing, invalid, errors, digests = evaluate_artifacts(
+            session_path,
+            expected,
+            expected_packet_digest,
+            findings_schema,
+            current_packet,
         )
         if len(done) == len(expected):
-            print(json.dumps({
+            if packet is not None:
+                final_packet, packet_error = reload_worker_packet(
+                    session_path,
+                    expected_packet_digest,
+                    require_admission=require_admission,
+                    legacy_v4=legacy_v4,
+                )
+                if packet_error:
+                    print(json.dumps({
+                        "status": "incomplete",
+                        "artifacts_done": [],
+                        "artifacts_missing": [],
+                        "artifacts_invalid": list(expected),
+                        "errors": {"WORKER-PACKET": packet_error},
+                        "count": 0,
+                    }, sort_keys=True))
+                    return False
+                admission_error = validate_pinned_admission(
+                    final_packet,
+                    pinned_admission,
+                )
+                if admission_error:
+                    print(json.dumps({
+                        "status": "incomplete",
+                        "artifacts_done": [],
+                        "artifacts_missing": [],
+                        "artifacts_invalid": list(expected),
+                        "errors": {"ADMISSION": admission_error},
+                        "count": 0,
+                    }, sort_keys=True))
+                    return False
+            output = {
                 "status": "complete",
                 "artifacts_done": done,
+                "artifact_digests": digests,
                 "artifacts_missing": [],
                 "artifacts_invalid": [],
                 "errors": {},
                 "count": len(done),
-            }, sort_keys=True))
+            }
+            if pinned_admission is not None:
+                output.update(pinned_admission)
+            print(json.dumps(output, sort_keys=True))
             return True
         if time.monotonic() >= deadline:
             print(json.dumps({
@@ -199,20 +2323,101 @@ def expected_overall_verdict(data) -> str:
     return "APPROVE"
 
 
-def validate_summary(data, expected_packet_digest: str):
-    if data.get("$schema") != SUMMARY_SCHEMA:
-        return f"$schema must be {SUMMARY_SCHEMA}"
+def load_completed_specialists(
+    session_path: Path,
+    packet,
+    completed,
+    expected_packet_digest: str,
+):
+    union = []
+    refs_by_axis = {axis: [] for axis in AXES}
+    for worker in packet["workers"]:
+        if worker["agent"] not in completed:
+            continue
+        data, _snapshot, error, is_missing = stable_read_repo_json(
+            packet["repo_root"],
+            worker["output"],
+            f"completed specialist {worker['agent']}",
+        )
+        if is_missing:
+            error = "artifact is missing"
+        if error is None:
+            error = validate_specialist(
+                data,
+                expected_packet_digest,
+                FINDINGS_SCHEMA,
+                worker,
+                packet,
+            )
+        if error:
+            return None, None, (
+                f"completed specialist {worker['agent']} is invalid: {error}"
+            )
+        refs_by_axis[worker["axis"]].append(worker["filename"])
+        union.extend(data["findings"])
+    return union, refs_by_axis, None
+
+
+def expected_axis_verdict(axis, union, packet, failed) -> str:
+    selected = [worker for worker in packet["workers"] if worker["axis"] == axis]
+    if not selected or any(worker["agent"] in failed for worker in selected):
+        return "INCOMPLETE"
+    severities = {
+        finding["severity"] for finding in union if finding.get("axis") == axis
+    }
+    return "FAIL" if severities.intersection({"P0", "P1"}) else "PASS"
+
+
+def validate_summary(
+    data,
+    expected_packet_digest: str,
+    summary_schema: str,
+    findings_schema: str,
+    session_path=None,
+    packet=None,
+):
+    if data.get("$schema") != summary_schema:
+        return f"$schema must be {summary_schema}"
     if data.get("mode") not in REVIEW_MODES:
         return "mode must be task, change, or plan"
-    for field in ["session", "change_id", "head", "context_digest"]:
+    if data.get("execution_mode") not in EXECUTION_MODES:
+        return "execution_mode must be isolated or sequential-shared-context"
+    for field in ["session", "change_id", "head"]:
         if not nonempty_string(data.get(field)):
             return f"{field} must be a non-empty string"
     if not digest_string(data.get("packet_digest")):
         return "packet_digest must be a lowercase SHA-256 digest"
     if data["packet_digest"] != expected_packet_digest:
         return f"packet_digest does not match expected immutable packet {expected_packet_digest}"
-    if data.get("worktree_digest") is not None and not nonempty_string(data.get("worktree_digest")):
-        return "worktree_digest must be null or a non-empty string"
+    admission_error = validate_artifact_admission_binding(data, packet, "SUMMARY.json")
+    if admission_error:
+        return admission_error
+    if packet is not None:
+        packet_data = packet["data"]
+        exact_fields = ["mode", "session", "change_id", "task_ids", "head"]
+        for field in exact_fields:
+            if data.get(field) != packet_data.get(field):
+                return f"summary {field} must match WORKER-PACKET.json"
+    if summary_schema == SUMMARY_SCHEMA:
+        if "worktree_digest" not in data or data.get("worktree_digest") is not None:
+            return (
+                "worktree_digest must be present and null until an exact worktree "
+                "observation is bound by the immutable Worker Packet"
+            )
+        if not digest_string(data.get("context_digest")):
+            return "context_digest must be a lowercase SHA-256 digest"
+        if packet is not None and data["context_digest"] != packet["context_digest"]:
+            return (
+                "context_digest must equal the immutable Worker Packet context digest "
+                f"{packet['context_digest']}"
+            )
+    else:
+        if not nonempty_string(data.get("context_digest")):
+            return "context_digest must be a non-empty string"
+        if data.get("worktree_digest") is not None and not nonempty_string(
+            data.get("worktree_digest")
+        ):
+            return "worktree_digest must be null or a non-empty string"
     task_ids = data.get("task_ids")
     if not isinstance(task_ids, list) or not all(nonempty_string(item) for item in task_ids):
         return "task_ids must be a string array"
@@ -228,8 +2433,10 @@ def validate_summary(data, expected_packet_digest: str):
         if not isinstance(item, dict) or item.get("verdict") not in AXIS_VERDICTS:
             return f"invalid {axis} verdict"
         refs = item.get("evidence_refs")
-        if not isinstance(refs, list) or not refs or not all(nonempty_string(ref) for ref in refs):
-            return f"{axis}.evidence_refs must be a non-empty string array"
+        if not isinstance(refs, list) or not all(nonempty_string(ref) for ref in refs):
+            return f"{axis}.evidence_refs must be a string array"
+        if packet is None and not refs:
+            return f"{axis}.evidence_refs must be non-empty for legacy summaries"
     workers = data.get("workers")
     if not isinstance(workers, dict):
         return "workers must be an object"
@@ -239,6 +2446,11 @@ def validate_summary(data, expected_packet_digest: str):
             return f"workers.{field} must be a string array"
         if len(values) != len(set(values)):
             return f"workers.{field} must not contain duplicates"
+    completed_set = set(workers["completed"])
+    failed_set = set(workers["failed"])
+    skipped_set = set(workers["skipped"])
+    if completed_set & failed_set or completed_set & skipped_set or failed_set & skipped_set:
+        return "workers completed, failed, and skipped sets must be disjoint"
     worker_selection = data.get("worker_selection")
     if not isinstance(worker_selection, list) or not worker_selection:
         return "worker_selection must be a non-empty array"
@@ -258,16 +2470,22 @@ def validate_summary(data, expected_packet_digest: str):
             return f"worker selection rationale is required: {worker}"
         seen_workers.add(worker)
         (selected if status == "selected" else skipped).add(worker)
-    if selected != set(workers["completed"]) | set(workers["failed"]):
+    if selected != completed_set | failed_set:
         return "selected workers must equal completed and failed workers"
-    if skipped != set(workers["skipped"]):
+    if skipped != skipped_set:
         return "skipped worker selection must match workers.skipped"
     if seen_workers != REVIEW_WORKERS:
         missing = sorted(REVIEW_WORKERS - seen_workers)
         extra = sorted(seen_workers - REVIEW_WORKERS)
         return f"worker_selection must disposition the complete roster; missing={missing}, extra={extra}"
-    if "review-spec" not in selected or "review-spec" not in set(workers["completed"]):
-        return "review-spec must be selected and completed"
+    if "review-spec" not in selected:
+        return "review-spec must be selected"
+    if packet is not None:
+        packet_selected = {worker["agent"] for worker in packet["workers"]}
+        if selected != packet_selected:
+            return "selected worker roster must equal WORKER-PACKET.json workers"
+        if skipped != REVIEW_WORKERS - packet_selected:
+            return "skipped worker roster must equal workers absent from WORKER-PACKET.json"
     findings = data.get("findings")
     if not isinstance(findings, list):
         return "findings must be an array"
@@ -276,44 +2494,351 @@ def validate_summary(data, expected_packet_digest: str):
         axis = finding.get("axis") if isinstance(finding, dict) else None
         if axis not in AXES:
             return "summary finding has invalid axis"
-        error = validate_finding(finding, axis, seen_ids)
+        error = validate_finding(
+            finding,
+            axis,
+            seen_ids,
+            findings_schema,
+            packet["trace"] if packet is not None else None,
+            packet["diff_scope"] if packet is not None else None,
+            packet["repo_root"] if packet is not None else None,
+        )
         if error:
             return error
     for field in ["positive_observations", "limitations"]:
         if not isinstance(data.get(field), list):
             return f"{field} must be an array"
+    coverage_refs = data.get("coverage_refs")
+    if not isinstance(coverage_refs, list) or not all(
+        nonempty_string(item) for item in coverage_refs
+    ):
+        return "coverage_refs must be a string array"
+    if not findings and not coverage_refs and not data["limitations"]:
+        return "zero findings require coverage_refs or an explicit limitation"
+    if packet is not None:
+        if session_path is None:
+            return "v4 summary validation requires the session path"
+        union, refs_by_axis, specialist_error = load_completed_specialists(
+            session_path,
+            packet,
+            set(workers["completed"]),
+            expected_packet_digest,
+        )
+        if specialist_error:
+            return specialist_error
+        if findings != union:
+            return (
+                "SUMMARY findings must equal the complete ordered specialist finding "
+                "union unchanged"
+            )
+        failed = set(workers["failed"])
+        for axis in AXES:
+            if axes[axis]["evidence_refs"] != refs_by_axis[axis]:
+                return f"{axis}.evidence_refs must equal packet-ordered completed specialist outputs"
+            axis_expected = expected_axis_verdict(axis, union, packet, failed)
+            if axes[axis]["verdict"] != axis_expected:
+                return (
+                    f"{axis} verdict {axes[axis]['verdict']} conflicts with completed "
+                    f"specialist evidence; expected {axis_expected}"
+                )
     expected = expected_overall_verdict(data)
     if data["verdict"] != expected:
         return f"verdict {data['verdict']} conflicts with evidence; expected {expected}"
     return None
 
 
-def wait_for_summary(session_path: Path, expected_packet_digest: str, timeout: float) -> bool:
-    summary_path = session_path / "SUMMARY.json"
+def wait_for_summary(
+    session_path: Path,
+    expected_packet_digest: str,
+    summary_schema: str,
+    findings_schema: str,
+    timeout: float,
+    summary_snapshot_bytes=None,
+    expected_summary_digest=None,
+    require_admission=True,
+    legacy_v4=False,
+) -> bool:
+    packet = None
+    if summary_schema == SUMMARY_SCHEMA:
+        packet, packet_error = reload_worker_packet(
+            session_path,
+            expected_packet_digest,
+            require_admission=require_admission,
+            legacy_v4=legacy_v4,
+        )
+        if packet_error:
+            print(json.dumps({"status": "incomplete", "error": packet_error}, sort_keys=True))
+            return False
+    pinned_admission = packet_admission_binding(packet)
+    repo_root = packet["repo_root"] if packet is not None else canonical_repo_root(session_path)[0]
+    summary_repo_path = f".ultra/reviews/{session_path.resolve().name}/SUMMARY.json"
     deadline = time.monotonic() + timeout
     last_error = "SUMMARY.json is missing"
     while True:
-        if summary_path.exists():
-            data, error = read_json(summary_path)
-            if error is None:
-                error = validate_summary(data, expected_packet_digest)
-            if error is None:
-                counts = {severity: 0 for severity in sorted(SEVERITIES)}
-                for finding in data["findings"]:
-                    counts[finding["severity"]] += 1
-                print(
+        current_packet = packet
+        if packet is not None:
+            current_packet, packet_error = reload_worker_packet(
+                session_path,
+                expected_packet_digest,
+                require_admission=require_admission,
+                legacy_v4=legacy_v4,
+            )
+            if packet_error:
+                print(json.dumps({"status": "incomplete", "error": packet_error}, sort_keys=True))
+                return False
+            admission_error = validate_pinned_admission(
+                current_packet,
+                pinned_admission,
+            )
+            if admission_error:
+                print(json.dumps({"status": "incomplete", "error": admission_error}, sort_keys=True))
+                return False
+
+        if summary_snapshot_bytes is not None:
+            summary_digest = hashlib.sha256(summary_snapshot_bytes).hexdigest()
+            if summary_digest != expected_summary_digest:
+                error = (
+                    "caller-owned SUMMARY snapshot bytes do not match "
+                    f"--summary-snapshot-digest {expected_summary_digest}; observed "
+                    f"{summary_digest}"
+                )
+                data = None
+            else:
+                data, error = decode_json_bytes(
+                    summary_snapshot_bytes,
+                    "caller-owned SUMMARY snapshot",
+                )
+            is_missing = False
+        else:
+            data, snapshot, error, is_missing = stable_read_repo_json(
+                repo_root,
+                summary_repo_path,
+                "SUMMARY.json",
+            )
+            summary_digest = snapshot["sha256"] if snapshot is not None else None
+
+        if error is None:
+            error = validate_summary(
+                data,
+                expected_packet_digest,
+                summary_schema,
+                findings_schema,
+                session_path,
+                current_packet,
+            )
+        if error is None:
+            if packet is not None:
+                final_packet, packet_error = reload_worker_packet(
+                    session_path,
+                    expected_packet_digest,
+                    require_admission=require_admission,
+                    legacy_v4=legacy_v4,
+                )
+                if packet_error:
+                    print(json.dumps({"status": "incomplete", "error": packet_error}, sort_keys=True))
+                    return False
+                admission_error = validate_pinned_admission(
+                    final_packet,
+                    pinned_admission,
+                )
+                if admission_error:
+                    print(json.dumps({"status": "incomplete", "error": admission_error}, sort_keys=True))
+                    return False
+            counts = {severity: 0 for severity in sorted(SEVERITIES)}
+            for finding in data["findings"]:
+                counts[finding["severity"]] += 1
+            output = {
+                "status": "complete",
+                "verdict": data["verdict"],
+                "message": (
                     f"Review complete: {data['verdict']} "
                     f"(P0:{counts['P0']} P1:{counts['P1']} P2:{counts['P2']} "
                     f"P3:{counts['P3']} total:{len(data['findings'])})"
-                )
-                return True
+                ),
+                "counts": {
+                    **counts,
+                    "total": len(data["findings"]),
+                },
+                "summary_digest": summary_digest,
+            }
+            if pinned_admission is not None:
+                output.update(pinned_admission)
+            print(json.dumps(output, sort_keys=True))
+            return True
+        if not is_missing:
             last_error = error
+        if summary_snapshot_bytes is not None:
+            print(json.dumps({"status": "incomplete", "error": last_error}, sort_keys=True))
+            return False
         if time.monotonic() >= deadline:
             print(json.dumps({"status": "incomplete", "error": last_error}, sort_keys=True))
             return False
         time.sleep(min(POLL_INTERVAL, max(0.0, deadline - time.monotonic())))
 
 
+def validate_packet_preflight(session_path: Path, expected_packet_digest: str) -> bool:
+    packet, error = reload_worker_packet(
+        session_path,
+        expected_packet_digest,
+    )
+    if error:
+        print(json.dumps({"status": "incomplete", "error": error}, sort_keys=True))
+        return False
+    admission_contract = packet["data"].get("admission_contract")
+    if admission_contract == ADMISSION_CONTRACT_V1:
+        retained, receipt_error = reload_worker_packet(
+            session_path,
+            expected_packet_digest,
+            require_admission=True,
+        )
+        if receipt_error:
+            print(json.dumps({
+                "status": "incomplete",
+                "error_code": "admission_v1_receipt_required",
+                "error": (
+                    "retained strict v1 is read-only and requires its exact existing "
+                    "ADMISSION.json receipt; start a fresh strict v2 review session "
+                    f"instead of creating or recreating it: {receipt_error}"
+                ),
+            }, sort_keys=True))
+            return False
+        binding = packet_admission_binding(retained)
+        print(json.dumps({
+            "status": "complete",
+            "packet_digest": retained["digest"],
+            "context_digest": retained["context_digest"],
+            "subject_digest": binding["subject_digest"],
+            "admission_digest": binding["admission_digest"],
+            "publication": "existing",
+        }, sort_keys=True))
+        return True
+    if admission_contract != ADMISSION_CONTRACT_V2:
+        print(json.dumps({
+            "status": "incomplete",
+            "error_code": "admission_contract_required",
+            "error": (
+                "WORKER-PACKET.json admission_contract must be "
+                f"{ADMISSION_CONTRACT_V2} for new strict v4 admission"
+            ),
+        }, sort_keys=True))
+        return False
+
+    admission_path = session_path / "ADMISSION.json"
+    try:
+        os.lstat(admission_path)
+        admission_missing = False
+    except FileNotFoundError:
+        admission_missing = True
+    except OSError as observation_error:
+        print(json.dumps({
+            "status": "incomplete",
+            "error_code": "admission_receipt_observation_failed",
+            "error": f"ADMISSION.json presence could not be observed safely: {observation_error}",
+        }, sort_keys=True))
+        return False
+    if admission_missing:
+        existing_outputs = []
+        output_paths = [
+            *(worker["output"] for worker in packet["workers"]),
+            f".ultra/reviews/{packet['data']['session']}/SUMMARY.json",
+        ]
+        for output_path in output_paths:
+            try:
+                os.lstat(packet["repo_root"] / output_path)
+                existing_outputs.append(output_path)
+            except FileNotFoundError:
+                continue
+            except OSError as observation_error:
+                print(json.dumps({
+                    "status": "incomplete",
+                    "error_code": "admission_output_observation_failed",
+                    "error": (
+                        f"review output presence could not be observed safely for "
+                        f"{output_path}: {observation_error}"
+                    ),
+                }, sort_keys=True))
+                return False
+        if existing_outputs:
+            print(json.dumps({
+                "status": "incomplete",
+                "error_code": "admission_receipt_missing_after_outputs",
+                "error": (
+                    "ADMISSION.json is missing after review output exists; the current "
+                    "strict session is INCOMPLETE and requires a fresh review session: "
+                    + ", ".join(existing_outputs)
+                ),
+            }, sort_keys=True))
+            return False
+
+    packet, error = reload_worker_packet(
+        session_path,
+        expected_packet_digest,
+        admit_subjects=True,
+    )
+    if error:
+        print(json.dumps({"status": "incomplete", "error": error}, sort_keys=True))
+        return False
+    publication, error = atomic_write_admission(
+        session_path,
+        packet["admission"],
+    )
+    if error:
+        print(json.dumps({
+            "status": "incomplete",
+            "error_code": error["code"],
+            "error": error["message"],
+        }, sort_keys=True))
+        return False
+    output = {
+        "status": "complete",
+        "packet_digest": packet["digest"],
+        "context_digest": packet["context_digest"],
+        "subject_digest": packet["admission"]["subject_digest"],
+        "admission_digest": publication["digest"],
+        "publication": publication["publication"],
+    }
+    if "durability_warning" in publication:
+        output["durability_warning"] = publication["durability_warning"]
+    print(json.dumps(output, sort_keys=True))
+    return True
+
+
+def validate_legacy_packet_boundary(
+    session_path: Path,
+    expected_packet_digest: str,
+    legacy_flag: str,
+):
+    repo_root, root_error = canonical_repo_root(session_path)
+    if root_error:
+        return f"WORKER-PACKET.json {root_error}"
+    packet_repo_path = (
+        f".ultra/reviews/{session_path.resolve().name}/WORKER-PACKET.json"
+    )
+    snapshot, error, missing = stable_read_repo_bytes(
+        repo_root,
+        packet_repo_path,
+        "WORKER-PACKET.json",
+    )
+    if missing:
+        return None
+    if error:
+        return error
+    if snapshot["sha256"] != expected_packet_digest:
+        return (
+            "WORKER-PACKET.json bytes do not match expected immutable packet digest "
+            f"{expected_packet_digest}; observed {snapshot['sha256']}"
+        )
+    packet, decode_error = decode_json_bytes(snapshot["bytes"], "WORKER-PACKET.json")
+    if decode_error:
+        return decode_error
+    if "admission_contract" in packet:
+        return (
+            "WORKER-PACKET.json declares a strict admission_contract and cannot be "
+            f"consumed with {legacy_flag}; retain and use its exact ADMISSION.json receipt"
+        )
+    return None
+
+
 def main():
     if len(sys.argv) < 3:
         print(__doc__, file=sys.stderr)
@@ -329,22 +2854,87 @@ def main():
         print("Error: --packet-digest requires the expected lowercase SHA-256 digest", file=sys.stderr)
         sys.exit(2)
     expected_packet_digest = sys.argv[4]
+    arguments = sys.argv[5:]
+    legacy_v3 = bool(arguments and arguments[0] == "--legacy-v3")
+    legacy_v4 = bool(arguments and arguments[0] == "--legacy-v4")
+    if legacy_v3 or legacy_v4:
+        arguments = arguments[1:]
+    findings_schema = LEGACY_FINDINGS_SCHEMA if legacy_v3 else FINDINGS_SCHEMA
+    summary_schema = LEGACY_SUMMARY_SCHEMA if legacy_v3 else SUMMARY_SCHEMA
+    require_admission = not legacy_v3 and not legacy_v4
+
+    if mode in {"agents", "summary"} and (legacy_v3 or legacy_v4):
+        legacy_flag = "--legacy-v3" if legacy_v3 else "--legacy-v4"
+        legacy_error = validate_legacy_packet_boundary(
+            session_path,
+            expected_packet_digest,
+            legacy_flag,
+        )
+        if legacy_error:
+            print(json.dumps({"status": "incomplete", "error": legacy_error}, sort_keys=True))
+            sys.exit(1)
 
-    if mode == "agents":
-        expected = sys.argv[5:]
+    if mode == "packet":
+        if legacy_v3 or legacy_v4 or arguments:
+            print("Error: packet mode accepts only --packet-digest", file=sys.stderr)
+            sys.exit(2)
+        ok = validate_packet_preflight(session_path, expected_packet_digest)
+    elif mode == "agents":
+        expected = arguments
         if not expected or len(expected) != len(set(expected)) or not all(
             ARTIFACT_STEM.fullmatch(stem) for stem in expected
         ):
             print("Error: agents mode requires unique safe artifact stems", file=sys.stderr)
             sys.exit(2)
-        ok = wait_for_agents(session_path, expected, expected_packet_digest, DEFAULT_TIMEOUT)
+        ok = wait_for_agents(
+            session_path,
+            expected,
+            expected_packet_digest,
+            findings_schema,
+            DEFAULT_TIMEOUT,
+            require_admission,
+            legacy_v4,
+        )
     elif mode == "summary":
-        if len(sys.argv) != 5:
-            print("Error: summary mode accepts only --packet-digest", file=sys.stderr)
-            sys.exit(2)
-        ok = wait_for_summary(session_path, expected_packet_digest, DEFAULT_TIMEOUT)
+        summary_snapshot_bytes = None
+        expected_summary_digest = None
+        if arguments:
+            if (
+                legacy_v3
+                or len(arguments) != 2
+                or arguments[0] != "--summary-snapshot-digest"
+                or not digest_string(arguments[1])
+            ):
+                print(
+                    "Error: summary mode accepts optional --legacy-v3, --legacy-v4, or "
+                    "--summary-snapshot-digest <sha256>",
+                    file=sys.stderr,
+                )
+                sys.exit(2)
+            expected_summary_digest = arguments[1]
+            summary_snapshot_bytes = sys.stdin.buffer.read(MAX_JSON_BYTES + 1)
+            if len(summary_snapshot_bytes) > MAX_JSON_BYTES:
+                print(json.dumps({
+                    "status": "incomplete",
+                    "error": (
+                        "caller-owned SUMMARY snapshot exceeds the "
+                        f"{MAX_JSON_BYTES}-byte artifact limit"
+                    ),
+                }, sort_keys=True))
+                sys.exit(1)
+        ok = wait_for_summary(
+            session_path,
+            expected_packet_digest,
+            summary_schema,
+            findings_schema,
+            DEFAULT_TIMEOUT,
+            summary_snapshot_bytes,
+            expected_summary_digest,
+            require_admission,
+            legacy_v4,
+        )
     else:
-        print(f"Error: unknown mode '{mode}'. Use 'agents' or 'summary'.", file=sys.stderr)
+        print(f"Error: unknown mode '{mode}'. Use 'packet', 'agents', or 'summary'.", file=sys.stderr)
         sys.exit(2)
     sys.exit(0 if ok else 1)