feat: add official Codex skill and harden task execution

Install and update artemis-use with task diagnostics and plugin guidance. Preserve cross-device file receipts, review canonical reports, guard stage transitions and reduce small history summaries. Improve exploration lifecycle, request recovery and browser live connections.

tests/unit/test_codex_skill_sync.py

@@ -0,0 +1,124 @@
+import json
+from pathlib import Path
+from unittest.mock import Mock
+
+import pytest
+
+from artemis.integrations import codex_skill as integration
+
+
+@pytest.fixture
+def bundle(tmp_path, monkeypatch):
+    root = tmp_path / "source"
+    source = root / "skills" / "artemis-use"
+    source.mkdir(parents=True)
+    (source / "SKILL.md").write_text(
+        "---\nname: artemis-use\ndescription: official\n---\n", "utf-8"
+    )
+    (source / "references").mkdir()
+    (source / "references" / "modes.md").write_text("official modes", "utf-8")
+    codex_home = tmp_path / "用户配置" / ".codex"
+    monkeypatch.setenv("CODEX_HOME", str(codex_home))
+    return root, source, codex_home
+
+
+def test_install_overwrites_entire_reserved_directory_and_is_noop_when_unchanged(bundle):
+    root, source, home = bundle
+    target = home / "skills" / "artemis-use"
+    target.mkdir(parents=True)
+    (target / "SKILL.md").write_text("unmanaged custom content", "utf-8")
+    (target / "custom.txt").write_text("must be removed", "utf-8")
+    unrelated = home / "skills" / "personal" / "SKILL.md"
+    unrelated.parent.mkdir()
+    unrelated.write_text("personal", "utf-8")
+    assert integration.sync_official_skill(root)["status"] == "updated"
+    assert not (target / "custom.txt").exists()
+    assert (target / "SKILL.md").read_bytes() == (source / "SKILL.md").read_bytes()
+    assert unrelated.read_text("utf-8") == "personal"
+    before = (target / "SKILL.md").stat().st_mtime_ns
+    assert integration.sync_official_skill(root)["status"] == "unchanged"
+    assert (target / "SKILL.md").stat().st_mtime_ns == before
+    (target / "SKILL.md").write_text("edited", "utf-8")
+    assert integration.sync_official_skill(root)["status"] == "updated"
+    assert (target / "SKILL.md").read_bytes() == (source / "SKILL.md").read_bytes()
+
+
+def test_update_removes_retired_files_and_migrates_only_marked_rules(bundle):
+    root, source, home = bundle
+    integration.sync_official_skill(root)
+    original = "user prefix\n<!-- BEGIN ARTEMIS MOBILE TESTING RULES -->\nold\n<!-- END ARTEMIS MOBILE TESTING RULES -->\nuser suffix\n"
+    for name in ("AGENTS.md", "AGENTS.override.md"):
+        (home / name).write_text(original, "utf-8")
+    (source / "references" / "modes.md").unlink()
+    integration.sync_official_skill(root)
+    assert not (home / "skills/artemis-use/references/modes.md").exists()
+    for name in ("AGENTS.md", "AGENTS.override.md"):
+        assert (home / name).read_text("utf-8") == "user prefix\nuser suffix\n"
+        assert (home / (name + ".artemis-update.bak")).read_text("utf-8") == original
+
+
+def test_failed_replacement_restores_previous_bundle(bundle, monkeypatch):
+    root, source, home = bundle
+    integration.sync_official_skill(root)
+    target = home / "skills/artemis-use"
+    original = (target / "SKILL.md").read_bytes()
+    (source / "SKILL.md").write_text("new version", "utf-8")
+    rename = Path.rename
+
+    def fail_staging(path, destination):
+        if path.name.startswith(".artemis-use-stage-"):
+            raise PermissionError("held open")
+        return rename(path, destination)
+
+    monkeypatch.setattr(Path, "rename", fail_staging)
+    assert integration.sync_skill_safely(root)["status"] == "error"
+    assert (target / "SKILL.md").read_bytes() == original
+    assert not list(target.parent.glob(".artemis-use-stage-*"))
+
+
+def test_newer_official_bundle_is_not_downgraded(bundle):
+    root, _, home = bundle
+    integration.sync_official_skill(root)
+    stamp = home / "skills/artemis-use" / integration.STAMP
+    data = json.loads(stamp.read_text("utf-8"))
+    data["artemis_version"] = "99.0.0"
+    stamp.write_text(json.dumps(data), "utf-8")
+    assert integration.sync_official_skill(root)["status"] == "newer_installed"
+
+
+def test_default_directory_and_duplicate_cleanup(bundle, tmp_path, monkeypatch):
+    root, source, _ = bundle
+    monkeypatch.delenv("CODEX_HOME")
+    monkeypatch.setattr(Path, "home", lambda: tmp_path)
+    legacy = tmp_path / ".codex/skills/artemis-use"
+    legacy.mkdir(parents=True)
+    (legacy / "SKILL.md").write_text("old", "utf-8")
+    assert integration.sync_official_skill(root)["path"] == str(legacy)
+    modern = tmp_path / ".agents/skills/artemis-use"
+    modern.mkdir(parents=True)
+    (modern / "SKILL.md").write_text("duplicate", "utf-8")
+    assert integration.sync_official_skill(root)["path"] == str(modern)
+    assert not legacy.exists()
+
+
+def test_actual_bundle_links_are_self_contained():
+    import re
+
+    source = Path(__file__).resolve().parents[2] / "skills/artemis-use"
+    for document in source.rglob("*.md"):
+        for relative in re.findall(r"\]\(([^)]+)\)", document.read_text("utf-8")):
+            if "://" in relative or relative.startswith("#"):
+                continue
+            target = (document.parent / relative.split("#", 1)[0]).resolve()
+            assert target.is_relative_to(source.resolve()), (document, relative)
+            assert target.exists(), (document, relative)
+
+
+def test_malformed_stamp_does_not_block_authoritative_update(bundle):
+    root, source, home = bundle
+    integration.sync_official_skill(root)
+    target = home / "skills/artemis-use"
+    (target / integration.STAMP).write_text("[]", "utf-8")
+    (target / "SKILL.md").write_text("unofficial", "utf-8")
+    assert integration.sync_official_skill(root)["status"] == "updated"
+    assert (target / "SKILL.md").read_bytes() == (source / "SKILL.md").read_bytes()