deerflow-code/offline-backend-20260512/backend/tests/test_bbc_page_generator_skill.py

from __future__ import annotations

import importlib.util
import json
from pathlib import Path

import pytest

from deerflow.skills.parser import parse_skill_file
from deerflow.skills.storage.local_skill_storage import LocalSkillStorage
from deerflow.skills.types import SkillCategory
from deerflow.skills.validation import _validate_skill_frontmatter

BACKEND_ROOT = Path(__file__).resolve().parents[1]
SKILL_ROOT = BACKEND_ROOT / "skills" / "public" / "bbc-page-generator"
SCRIPT_PATH = SKILL_ROOT / "scripts" / "render_bbc_page.py"
LEAD_PROMPT_PATH = (
    BACKEND_ROOT
    / "packages"
    / "harness"
    / "deerflow"
    / "agents"
    / "lead_agent"
    / "prompt.py"
)


def _load_renderer():
    spec = importlib.util.spec_from_file_location(
        "bbc_page_renderer",
        SCRIPT_PATH,
    )
    assert spec and spec.loader
    module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(module)
    return module


def test_skill_is_discoverable_and_valid() -> None:
    valid, error, _metadata = _validate_skill_frontmatter(SKILL_ROOT)
    assert valid, error
    skill = parse_skill_file(SKILL_ROOT / "SKILL.md", SkillCategory.PUBLIC)
    assert skill is not None
    assert skill.name == "bbc-page-generator"
    assert "BBC" in skill.description

    discovered = LocalSkillStorage(
        host_path=str(SKILL_ROOT.parent.parent)
    ).load_skills(enabled_only=True)
    discovered_by_name = {item.name: item for item in discovered}
    assert (
        discovered_by_name["bbc-page-generator"].category
        == SkillCategory.PUBLIC
    )
    assert discovered_by_name["bbc-page-generator"].enabled is True

    lead_prompt_source = LEAD_PROMPT_PATH.read_text(encoding="utf-8")
    assert "/mnt/skills/public/bbc-page-generator/SKILL.md" in lead_prompt_source
    assert "专用 BBC 路由" in lead_prompt_source


def test_default_json_renders_editable_bbc_html(tmp_path: Path) -> None:
    renderer = _load_renderer()
    data = json.loads(
        (SKILL_ROOT / "assets" / "default-data.json").read_text(
            encoding="utf-8"
        )
    )
    summary = renderer.validate_document(data)
    template = (SKILL_ROOT / "assets" / "bbc-template.html").read_text(
        encoding="utf-8"
    )
    runtime = (SKILL_ROOT / "assets" / "bbc-runtime.js").read_text(
        encoding="utf-8"
    )
    html = renderer.render_document(data, template, runtime)

    assert summary["modules"] == list(renderer.MODULES)
    assert summary["counts"]["focus"] >= 1
    assert renderer.DATA_MARKER not in html
    assert renderer.RUNTIME_MARKER not in html
    assert 'id="bbcPageData"' in html
    assert 'id="leadLayout"' in html
    assert 'id="mediaGrid"' in html
    assert 'id="audioPlay"' in html
    assert 'id="liveModal"' in html
    assert 'id="openSettings"' in html
    assert 'id="moduleTabs"' in html
    assert 'id="applyJson"' in html
    assert "uploadJsonMedia" in html
    assert "installInlineMediaReplace" in html
    assert "dfInlineMediaInput" in html
    assert 'input.accept = "image/*,video/*"' in html
    assert 'button.title = "上传图片或视频并替换"' in html
    assert "active.expected" not in html
    assert "previous.posterUrl" not in html
    assert "decorateBuiltInVisuals" in html
    assert 'data-video-id="' in html
    assert "replaceBuiltInVisuals" not in html
    assert "readAsDataURL" in html
    assert '"brandLetters": [' in html
    assert "<title>BBC 新闻 — 看见正在发生的世界</title>" in html
    assert '"bbc-page-data-v1:" + pageFingerprint' in html
    assert "const readStorage = (key)" in html
    assert "window.localStorage.getItem" in html
    assert "const cachedData = localStorage.getItem" not in html
    assert '<img src="http' not in html
    assert '<video src="http' not in html

    data["site"]["pageTitle"] = "BBC · 智能体公共新闻"
    titled_html = renderer.render_document(data, template, runtime)
    assert "<title>BBC · 智能体公共新闻</title>" in titled_html
    assert "<title>BBC 新闻 — 看见正在发生的世界</title>" not in titled_html

    data["lead"]["title"] = "</script><script>alert('x')</script>"
    escaped_html = renderer.render_document(data, template, runtime)
    assert "</script><script>alert" not in escaped_html
    assert "<\\/script><script>alert" in escaped_html

    output = tmp_path / "bbc-page.html"
    renderer._write_new(output, html)
    assert output.is_file()
    with pytest.raises(
        renderer.DataValidationError,
        match="Refusing to overwrite",
    ):
        renderer._write_new(output, html)


def test_validation_rejects_missing_duplicate_and_remote_visual() -> None:
    renderer = _load_renderer()
    data = json.loads(
        (SKILL_ROOT / "assets" / "default-data.json").read_text(
            encoding="utf-8"
        )
    )
    data.pop("footer")
    data["focus"]["items"][1]["id"] = data["focus"]["items"][0]["id"]
    data["focus"]["items"][0]["visual"] = "remote-image"

    with pytest.raises(renderer.DataValidationError) as exc_info:
        renderer.validate_document(data)

    message = str(exc_info.value)
    assert "Missing top-level module: footer" in message
    assert "must be unique" in message
    assert "visual must be one of" in message


def test_remote_media_contract_supports_image_video_and_webpage() -> None:
    renderer = _load_renderer()
    source = json.loads((SKILL_ROOT / "assets" / "default-data.json").read_text(encoding="utf-8"))
    for media in (
        {"type": "image", "mediaUrl": "https://cdn.example.com/news.jpg", "alt": "新闻图片"},
        {"type": "video", "mediaUrl": "https://cdn.example.com/news.mp4", "posterUrl": "https://cdn.example.com/poster.jpg"},
        {"type": "webpage", "pageUrl": "https://example.com/story", "title": "新闻原文", "embed": True},
    ):
        data = json.loads(json.dumps(source, ensure_ascii=False))
        data["lead"]["media"] = media
        renderer.validate_document(data)
    runtime = (SKILL_ROOT / "assets" / "bbc-runtime.js").read_text(encoding="utf-8")
    assert 'media.type === "webpage"' in runtime
    assert 'sandbox="allow-forms allow-popups allow-scripts"' in runtime