Gaming Night Brief

Research Execution Protocol (Mandatory)

• All web research must be executed via OpenClaw agent tools.
• OpenClaw web_search is the primary retrieval backend (current provider: Perplexity Sonar).
• Require deep search + cross-source verification before output.
• If web_search is unavailable, explicitly report collection failure and stop.

Perplexity Deep-Search Rules (Mandatory)

1. Time window gate:

• Delta window is past 12-24h by default.
• Older updates can be used as trend context only, not "tonight change".

2. Evidence density:

• Every top change should have at least 2 independent sources when feasible.
• Single-source items must be labeled 单源未完全证实.

3. Source metadata:

• Each cited item must include published_at, fetched_at, and URL.
• If publication time is unknown, mark 发布时间未知.

4. Conflict handling:

• If outlet claims conflict (e.g., launch timing, sales, DAU), include 分歧点 and avoid definitive numbers.
• Keep confidence conservative for rumor-heavy topics.

5. Fact vs inference split:

• Facts = source-backed updates.
• Inference = platform strategy interpretation and short-term impact guess, explicitly labeled.

Continuity protocol (mandatory)

• Compare against previous-night briefing:

1. What remained the same trend
2. What changed tonight
3. Which segment got hotter/colder

• If previous report is unavailable, state this explicitly.

Stage 1: Multi-platform source search

Search 18-30 sources; avoid single-platform bias. Prioritize:

1. Official announcements / publisher statements
2. Platform stores/rankings and event updates
3. Reputable game media and industry analysis
4. Community signal only as secondary evidence (not primary facts)

Stage 2: Signal extraction (internal)

• Separate real updates from PR noise.
• Pick top 3-5 meaningful deltas.
• For each delta, identify stakeholder impact:
  • Players
  • Developers/studios
  • Platforms/publishers

Stage 3: Telegram output (strict)

• No # headers.
• Use Bold + Emoji section headers.
• Use > for plain-language explanations.
• Keep sections compact and skimmable.
• Use relative-time citations.

Output Template

📡 游戏圈晚报工作量声明
本次检索约XX个来源，交叉验证XX条有效增量信息（重点覆盖过去12-24小时）。

🧱 事实层（可验证） • 事实1：...【来源A/B】
• 事实2：...【来源A/B】

🧠 推断层（带置信度） • 推断1（高/中/低）：...
• 推断2（高/中/低）：...

🎮 今晚最值得看的三条变化
• 变化1：一句话
• 变化2：一句话
• 变化3：一句话（可选）

🔁 昨晚 vs 今晚变化
• 延续趋势：...
• 新增变量：...
• 热度迁移：移动端 / 主机端 / PC端（谁升温，谁降温）

🧠 大白话拆解：为什么这条新闻会出现
事件A
• 表面新闻：...
• 本质驱动：商业策略 / 发行节奏 / 平台博弈 / 成本压力

大白话：把行业术语翻译成普通玩家也能懂的话。

🕹️ 分平台影响地图
• TapTap/手游：...
• 主机（PS5/Xbox/Nintendo）：...
• PC/Steam/Epic：...

📊 平台热度记分板（0-5） • 手游：X/5（依据：...）
• 主机：X/5（依据：...）
• PC平台：X/5（依据：...）

🧪 信息置信度标注 • 变化1：已证实 / 多源一致 / 单源未证实
• 变化2：...

⚖️ 分歧与不确定性 • 分歧点：...
• 当前处理：为何降低置信度/暂不下定论。

📅 前瞻清单（未来1-4周）
• 需盯的发布会/测试节点：...
• 可能引发二次波动的变量：...

✅ 明日验证点（防空话）

1. 若...发生，则今天判断被验证/被否证
2. 若...未发生，则下调结论置信度

🧭 睡前一页纸结论
• 今晚最重要一句话：...
• 明天值得继续追踪的信号：...

🔎 证据附录（3-8条）

1. [来源名称] 标题（发布时间：UTC...，检索时间：UTC...）- 用于支撑：...
2. ...

Quality Rules

1. Facts first; clearly label inference.
2. Avoid rumor-led conclusions unless clearly marked 未证实.
3. No fabricated release dates, sales numbers, or quote attribution.
4. If little changed, explicitly state 今晚游戏圈无重大增量.
5. Web evidence must follow OpenClaw web_search (Perplexity) execution path.
6. Do not fabricate workload counts; if exact count is unavailable, state 系统未返回精确统计.
7. Every top change must have at least one corresponding source in 证据附录.
8. Anything older than 24h must be marked as background, not tonight delta.