loopx-project · huangruiteng · Sep 15, 2026 · Sep 14, 2026 · Sep 14, 2026 · Sep 15, 2026
@@ -2928,6 +2928,11 @@ an equal-byte retry syncs file and directory before reporting `current`. Narrati
 canonical records stay intact. This converges the retained Python presentation/legacy
 input adapter; it adds no RPC or business state machine and does not change TS authority
 transactions, provider defaults, SQLite D2 or D3 promotion requirements.
+Objective examples are now isolated by both document producers, and display
+readback shares the Goal codec. Canonical Todo reads remain independent of
+malformed display; Objective narrative is not added to the Todo store or its
+recovery scope. This is a bounded D1 adapter correction, not D1–D3 qualification.
+See the [document boundary](../../reference/protocols/active-state-structured-projection-v0.md#markdown-ownership-boundary).
 Canonical handoff-mode show/set no longer depend on Markdown frontmatter or local lease files. One TS transaction binds quiescence, mode and durable operation replay to the same revision, including sealed no-op intents. This adds a provider-neutral command boundary, not a provider default or whole-Goal cutover; frontmatter remains outside the Todo-section renderer. See [operation and recovery](../../reference/handoff-mode.md).
 
 T2 now commits a lease-free native Monitor observation and its independent

@@ -2318,6 +2318,10 @@ fenced 示例被当成真实任务、归档 end marker 后叙述进入历史、
 这是永久 Python 展示／legacy 输入适配层的收敛：TS authority transaction、provider
 默认值、SQLite D2 与 D3 promotion 合同不变，不增加 RPC 或另一份业务状态机。
 Canonical handoff-mode show/set 不再依赖 Markdown frontmatter 或本地 lease；一笔 TS 事务把空闲检查、mode 与耐久操作回执绑定到同一 revision，包括未改值请求的回执。该命令边界不切换默认 provider、不晋升整 Goal；frontmatter 仍不属于 Todo-section renderer。操作与恢复见 [handoff-mode](../../reference/handoff-mode.md)。
+两个文档生成入口现在隔离 Objective 示例，展示读回复用 Goal codec。Canonical Todo
+读取仍独立于畸形展示；Objective 叙述不进入 Todo store，也不扩大其恢复范围。
+这是有界 D1 适配层修正，不代表 D1–D3 资格化。
+见[文档边界](../../reference/protocols/active-state-structured-projection-v0.md#markdown-ownership-boundary)。
 
 能力缺口 consumer 在 legacy/canonical 输入上共用 TS requirement/resolution owner，
 包括 quota 的 Monitor 能力分流。删除 Python missing-set 与 owner/repair 决策 builder，

@@ -397,6 +397,11 @@ owns durable truth, recovery, cutover, and projection delivery. Neither a fully
 TypeScript CLI nor `loopxd` is a prerequisite for removing Python decisions.
 An input adapter or external-effect executor may remain Python.
 
+Objective serialization and readback now share the existing Goal metadata and
+section owners; legacy registration compares decoded values and exact narrative.
+This closes a presentation boundary without adding TS transport or moving business
+authority. See the [document boundary](../../reference/protocols/active-state-structured-projection-v0.md#markdown-ownership-boundary).
+
 The lifecycle-admission slice now uses `todo_lifecycle_decision.ts` for legacy
 claim/update admission, delegated action/reason checks, ownership-holder routing,
 and native complete/supersede. Native text/note edits and terminal transitions

@@ -313,6 +313,10 @@ Markdown renderer 长期保留。
 负责 durable truth、恢复、cutover 与投影交付。删除 Python decision 不以前端 CLI
 全部改成 TypeScript 或 `loopxd` 落地为前提；输入适配和外部 effect 执行可以保留 Python。
 
+Objective 编码与读回复用既有 Goal metadata／section owner；旧注册按解码后的值与
+精确叙述正文比较。这是展示边界闭合，不增加 TS transport，也不迁移业务权威。
+见[文档边界](../../reference/protocols/active-state-structured-projection-v0.md#markdown-ownership-boundary)。
+
 本次 lifecycle-admission 切片将 legacy claim/update 准入、委托 action/reason 检查、
 ownership-holder 路由及 native complete/supersede 统一到
 `todo_lifecycle_decision.ts`。Native text/note 编辑与 terminal transition 在进程内

@@ -119,6 +119,25 @@ Directly editing a projection is not a state transition.
 
 ## Markdown Ownership Boundary
 
+New bootstrap and project-registration documents quote each Objective line and
+escape HTML metacharacters. Fences, comments, headings, and Todo markers in the
+objective remain content rather than document structure. Frontmatter string
+encoding and readback share JSON semantics, including escaped Unicode line
+separators; only complete delimiter lines terminate frontmatter. Objective
+readback composes the existing section reader and decodes generated quotation.
+Registration compares metadata values and exact remaining narrative, accepting
+legacy Objective presentation without rewriting it; changed content still conflicts.
+
+This is the permanent Python presentation/legacy-input adapter described by the
+[TypeScript RFC](../../architecture/rfcs/typescript-control-plane-migration-v0.md#next-delivery-sequence)
+and [shared-authority RFC](../../architecture/rfcs/shared-goal-authority-state-provider-v0.md#next-delivery-and-parallel-provider-work).
+It adds no business rule, RPC, provider, or authority write. Post-cutover Todo
+consumers still read canonical state when Markdown is absent or malformed;
+rendering never imports Objective examples into that state. Before cutover,
+the existing legacy writer remains subject to its normal fence. Objective is
+independent Goal narrative, outside the Todo store and Todo-section recovery.
+Existing malformed documents are not automatically repaired.
+
 Markdown is not one undifferentiated database row. Agents generate and maintain
 both its structured sections and narrative through LoopX. The distinction is
 canonical ownership, not human versus Agent authorship: after promotion,

@@ -40,6 +40,7 @@
     MULTI_SUBAGENT_ORCHESTRATION_MODE,
 )
 from .paths import rel_or_abs, resolve_runtime_root
+from .control_plane.goals.active_state_metadata import markdown_blockquote, markdown_frontmatter_string
 from .registry_writability import probe_registry_write_path
 from .todos import add_todo_to_lines
 
@@ -481,7 +482,7 @@ def render_state_markdown(
     include_connection_validation: bool = True,
     handoff_mode: str = HANDOFF_MODE_LEGACY,
 ) -> str:
-    safe_objective = objective.replace('"', '\\"')
+    safe_objective = markdown_frontmatter_string(objective)
     profile_summary = execution_profile_summary(execution_profile)
     onboarding_markdown = render_onboarding_state_markdown(
         onboarding_scan=onboarding_scan,
@@ -507,7 +508,7 @@ def render_state_markdown(
     state_text = f"""---
 status: active
 owner_mode: goal
-objective: "{safe_objective}"
+objective: {safe_objective}
 updated_at: {updated_at}
 adapter_id: {goal_id}
 {handoff_mode_line}---
@@ -516,7 +517,7 @@ def render_state_markdown(
 
 ## Objective
 
-{objective}
+{markdown_blockquote(objective)}
 
 ## Authority Sources
 

@@ -38,6 +38,7 @@
 from .chat_store import ChatSessionStore
 from .capabilities.manager_runtime import manager_runtime_capability_projection
 from .capabilities.manager_context.roundtrip import project_chat_session_snapshot
+from .control_plane.goals.active_state_metadata import active_state_section_text
 from .control_plane.status.ssh_host_catalog import (
     SSH_HOST_CATALOG_PATH,
     ssh_host_catalog_payload,
@@ -123,22 +124,6 @@ def _compact_text(value: Any, *, limit: int = 600) -> str:
     return " ".join(str(value or "").split())[:limit].strip()
 
 
-def _active_state_section(state_text: str, heading: str) -> str:
-    marker = f"## {heading}"
-    start = state_text.find(marker)
-    if start < 0:
-        return ""
-    content_start = start + len(marker)
-    end = state_text.find("\n## ", content_start)
-    section = state_text[content_start : end if end >= 0 else None]
-    lines = [
-        line.strip().removeprefix("- ").strip()
-        for line in section.splitlines()
-        if line.strip() and not line.lstrip().startswith("<!--")
-    ]
-    return _compact_text(" ".join(lines))
-
-
 def _goal_public_context(registry: dict[str, Any], goal: dict[str, Any]) -> dict[str, Any]:
     goal_id = str(goal.get("id") or "")
     project = Path(str(goal.get("repo") or ".")).expanduser().resolve()
@@ -148,7 +133,7 @@ def _goal_public_context(registry: dict[str, Any], goal: dict[str, Any]) -> dict
     if state_path is not None and state_path.exists():
         try:
             state_text = state_path.read_text(encoding="utf-8")
-            objective = _active_state_section(state_text, "Objective")
+            objective = _compact_text(active_state_section_text(state_text, "Objective"))
             title_line = next(
                 (line[2:].strip() for line in state_text.splitlines() if line.startswith("# ")),
                 "",

@@ -1,5 +1,11 @@
 from __future__ import annotations
 
+import json
+import re
+from html import escape, unescape
+
+from .active_state_sections import active_state_sections
+
 
 USER_TODO_HEADER_MARKERS = (
     "user todo",
@@ -21,19 +27,63 @@
 )
 
 
-def parse_state_frontmatter(state_text: str) -> dict[str, str]:
-    if not state_text.startswith("---"):
-        return {}
-    parts = state_text.split("---", 2)
-    if len(parts) < 3:
-        return {}
+def markdown_frontmatter_string(value: str) -> str:
+    encoded = json.dumps(value, ensure_ascii=False)
+    for separator in ("\x85", "\u2028", "\u2029"):
+        encoded = encoded.replace(separator, f"\\u{ord(separator):04x}")
+    return encoded
+
+
+def markdown_blockquote(value: str) -> str:
+    return "\n".join(f"> {escape(line, quote=False)}" for line in value.splitlines())
+
+
+def active_state_section_text(state_text: str, heading: str) -> str:
+    _, body = split_state_frontmatter(state_text)
+    sections = active_state_sections(
+        body, (heading,), section_heading_pattern=re.compile(r"^## (.+?)[ \t]*$"),
+    )
+    section_lines = [line for line in sections[heading] if line]
+    if heading == "Objective" and section_lines and all(
+        line.startswith("> ") for line in section_lines
+    ):
+        text = " ".join(unescape(line[2:]) for line in section_lines)
+    else:
+        text = " ".join(
+            line.strip().removeprefix("- ").strip()
+            for line in section_lines
+            if line.strip() and not line.lstrip().startswith("<!--")
+        )
+    return " ".join(text.split())
+
+
+def split_state_frontmatter(state_text: str) -> tuple[dict[str, str], str]:
+    """Decode generated string metadata; retain the legacy unquoted input form."""
+    match = re.match(r"\A---[ \t]*\r?\n(.*?)^---[ \t]*(?:\r?\n|\Z)", state_text, re.M | re.S)
+    if match is None:
+        return {}, state_text
     result: dict[str, str] = {}
-    for line in parts[1].splitlines():
+    # JSON strings may contain Unicode separators: only physical LF ends a field.
+    for line in match.group(1).split("\n"):
         if ":" not in line:
             continue
         key, value = line.split(":", 1)
-        result[key.strip()] = value.strip().strip('"')
-    return result
+        value = value.strip()
+        if value.startswith('"'):
+            try:
+                decoded = json.loads(value)
+            except json.JSONDecodeError:
+                pass  # Legacy writers did not escape backslashes as JSON.
+            else:
+                if isinstance(decoded, str):
+                    result[key.strip()] = decoded
+                    continue
+        result[key.strip()] = value.strip('"')
+    return result, state_text[match.end():]
+
+
+def parse_state_frontmatter(state_text: str) -> dict[str, str]:
+    return split_state_frontmatter(state_text)[0]
 
 
 def todo_role_for_heading(heading: str) -> str | None:

@@ -13,6 +13,9 @@
 from ..coordination.legacy_writer_fence import legacy_todo_write_transaction, require_legacy_state_replacement_allowed
 from ...paths import DEFAULT_RUNTIME_ROOT
 from ...registry import atomic_write_json
+from ..goals.active_state_metadata import (
+    markdown_blockquote, markdown_frontmatter_string, split_state_frontmatter,
+)
 from ...repository_identity import normalize_repository_identity
 from .contract import validate_project_record_bindings
 
@@ -133,7 +136,7 @@ def bullets(items: list[str], *, empty: str) -> str:
 status: active
 owner_mode: goal
 project_id: {json.dumps(project_id, ensure_ascii=False)}
-objective: {json.dumps(objective, ensure_ascii=False)}
+objective: {markdown_frontmatter_string(objective)}
 updated_at: {updated_at}
 adapter_id: {goal_id}
 ---
@@ -142,7 +145,7 @@ def bullets(items: list[str], *, empty: str) -> str:
 
 ## Objective
 
-{objective}
+{markdown_blockquote(objective)}
 
 ## Acceptance
 
@@ -174,6 +177,22 @@ def bullets(items: list[str], *, empty: str) -> str:
 """
 
 
+def _registration_state_matches(existing: str, expected: str, *, objective: str) -> bool:
+    """Compare metadata values and exact narrative without rewriting old state."""
+    existing_metadata, existing_body = split_state_frontmatter(existing)
+    expected_metadata, expected_body = split_state_frontmatter(expected)
+    if existing_metadata != expected_metadata:
+        return False
+    marker = "\n## Objective\n\n"
+    existing_prefix, separator, existing_section = existing_body.partition(marker)
+    expected_prefix, _, expected_section = expected_body.partition(marker)
+    if not separator or existing_prefix != expected_prefix:
+        return False
+    quoted = markdown_blockquote(objective)
+    remainder = expected_section[len(quoted):]
+    return existing_section in (quoted + remainder, objective + remainder)
+
+
 def register_project_goal(
     *,
     registry_path: Path,
@@ -326,7 +345,9 @@ def register_project_goal(
                     if existing_updated_at is not None
                     else None
                 )
-                if existing_state != matching_state:
+                if matching_state is None or not _registration_state_matches(
+                    existing_state, matching_state, objective=objective,
+                ):
                     raise ValueError(
                         f"goal state file conflicts with registration: {state_file}"
                     )