Implement WebAssembly runtime and kinematics coverage

scripts/validate_task_pointer.py

@@ -0,0 +1,89 @@
+#!/usr/bin/env python3
+"""Validate the small task pointer loaded at the start of a work session."""
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+from pathlib import Path
+
+
+MAX_POINTER_BYTES = 8192
+MAX_POINTER_LINES = 180
+REQUIRED_HEADINGS = ("## Current", "## Read", "## Task", "## Next", "## Verify", "## Stop")
+FORBIDDEN_HEADINGS = (
+    "## 6. 已验证证据",
+    "## Evidence",
+    "## History",
+    "## 历史",
+    "## Chat",
+)
+
+
+class PointerError(ValueError):
+    """Raised when the bootstrap pointer would expand the model context."""
+
+
+def validate_pointer(
+    path: Path,
+    *,
+    max_bytes: int = MAX_POINTER_BYTES,
+    max_lines: int = MAX_POINTER_LINES,
+    lock_path: Path | None = None,
+) -> dict[str, int]:
+    text = path.read_text(encoding="utf-8")
+    size = len(text.encode("utf-8"))
+    lines = len(text.splitlines())
+    if size > max_bytes:
+        raise PointerError(f"task pointer exceeds byte budget ({size}>{max_bytes})")
+    if lines > max_lines:
+        raise PointerError(f"task pointer exceeds line budget ({lines}>{max_lines})")
+    missing = [heading for heading in REQUIRED_HEADINGS if heading not in text]
+    if missing:
+        raise PointerError(f"task pointer missing sections: {', '.join(missing)}")
+    found = [heading for heading in FORBIDDEN_HEADINGS if re.search(rf"^\s*{re.escape(heading)}\s*$", text, re.MULTILINE)]
+    if found:
+        raise PointerError(f"task pointer contains history section: {found[0]}")
+    if "coverage/" not in text:
+        raise PointerError("task pointer must direct large evidence to paths")
+    pointer_match = re.search(r"^context_id:\s*([^\s]+)\s*$", text, re.MULTILINE)
+    if pointer_match is None:
+        raise PointerError("task pointer is missing context_id")
+    if lock_path is not None and lock_path.exists():
+        try:
+            lock = json.loads(lock_path.read_text(encoding="utf-8"))
+        except (OSError, UnicodeError, json.JSONDecodeError) as exc:
+            raise PointerError(f"cannot read context lock: {exc}") from exc
+        lock_id = lock.get("context_id") if isinstance(lock, dict) else None
+        if not lock_id:
+            raise PointerError("context lock is missing context_id")
+        if pointer_match.group(1) != str(lock_id):
+            raise PointerError(
+                f"task pointer context_id does not match lock ({pointer_match.group(1)}!={lock_id})"
+            )
+    return {"bytes": size, "lines": lines, "max_bytes": max_bytes, "max_lines": max_lines}
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("path", nargs="?", type=Path, default=Path("nextTask.md"))
+    parser.add_argument("--max-bytes", type=int, default=MAX_POINTER_BYTES)
+    parser.add_argument("--max-lines", type=int, default=MAX_POINTER_LINES)
+    parser.add_argument("--lock", type=Path, default=Path("context-lock.json"))
+    args = parser.parse_args()
+    try:
+        result = validate_pointer(
+            args.path,
+            max_bytes=args.max_bytes,
+            max_lines=args.max_lines,
+            lock_path=args.lock,
+        )
+    except (OSError, UnicodeError, PointerError) as exc:
+        parser.error(str(exc))
+    print(f"task_pointer=valid bytes={result['bytes']} lines={result['lines']}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())