video-script— agent skill

What this does

Authoring + validation of the narration script. The agent writes `work_dir/narration.json` following the rules below; then validate.py lints it against the understanding index, and in full mode time-aligns it to quiet windows.

Step 1 — read the brief

Read work_dir/agent_narration_brief.md (scenes, durations, quiet windows, char budget) first. Digest long dialogue via asr_writing_chunks.json; judge "is there speech/a silent slot here?" via timeline_fusion.json. Check raw vlm_analysis.json / asr_result.json for details. In full mode, timestamps are original-video time. In orchestrated cut mode, pass 1 only writes clip_plan.json; after edited_source.mp4 exists, pass 2 writes narration.json in output timeline time.

写稿前先跑 python3 skills/video-recap/scripts/recap_inspect.py --work-dir <work_dir> state 看清楚当前模式、缺哪个产物、下一步该写什么。cut pass 2 写解说时用 recap_inspect.py --work-dir <work_dir> clip-map --output-start <s> --output-end <e>（或 --source-start/--source-end）核对输出↔原片时间轴，确认某段成片对应哪段原片、有没有跨剪辑边界或落进被剪掉的区间。

Step 2 — write narration.json

[
  {"start": 5.0, "end": 12.0, "narration": "解说文本。", "pause_after_ms": 250, "overlaps_speech": true}
]

Optionally also author original_subtitles.json — [{start,end,text}] (OUTPUT time) — the calibrated original dialogue burned during the original-audio gaps (ASR errors/names fixed, only what is actually spoken there). Rendered in 「」 to set it apart from narration. If omitted, assemble falls back to a conservative auto-ASR mapping. See the brief's 原声留白字幕 section.

写作规则（BLOCK recap + 原声留白）

1. 按 BLOCK 写：每个 beat 是一个 2–4 句的连续想法，合成为一条流畅 TTS；不要回到“一句一停”的碎句模式。
2. 约 7:3 节奏：让解说覆盖约 70% 成片时间，刻意留下约 30% 多秒原声留白，让关键对白、动作或音乐完整呼吸。
3. 块与原声交替：一块解说铺垫原声留白，下一块承接原声刚呈现的信息；禁止墙到墙不停讲，也禁止每句之间机械小缝。
4. 按 brief 的 block 数量和有效语速控量：每块窗口按 字数 / brief 头部 speech budget 估算，太长就拆块，太短就合并成完整想法。
5. 不要看图说话：观众看得见动作表情，解说讲动机、关系、潜台词和剧情意义（基于画面可见的证据，不要编造）。
6. 用已知角色名：--context 或 background_research.json 给出角色名时优先使用。
7. 完整句子：以句号 / 问号 / 感叹号结束，不写半句。

解说手法（区分“解说”和“字幕”）

• 钩子：开头 1-2 个 beat 制造悬念或利害，不是交代场景。
• 主线：选一条主线（目标 / 关系 / 悬念），每个 beat 都推进它。
• 递进：信息和张力逐步升级。悬念缺口：提前埋后果，后面回收。
• 收尾：最后 1-2 个 beat 给出结果或反转，不要泛泛收场。
• 衔接：解说块和它旁边的原声留白是同一个 beat——块尾给原声铺垫，下一块承接原声刚呈现的内容，别和原声各说各的。
• 给信息而非念画面；去废词：用具体名词动词，删空泛形容。

Step 2.5 — review GATE (advisory, logged, overridable)

A separate quality pass (LLM-as-judge), distinct from the mechanical lint below. Needs the chat API key (same as VLM).

1. Run: python3 scripts/review.py --work-dir <work_dir> (auto-detects cut mode and grounds against the OUTPUT timeline when a validated cut is present; pass --timeline source to force source grounding)
2. Open narration_review.md. For every error finding (ESPECIALLY category=hallucination — a claim

not grounded in the visual/ASR evidence), revise narration.json and re-run review until either:

• (a) verdict == OK with zero error findings, OR
• (b) you consciously OVERRIDE a remaining finding (next step).

1. To OVERRIDE: append a block to work_dir/narration_review_override.md naming WHICH finding

(segment + category), WHY it is acceptable, and who signed off. Unaddressed error findings with no override entry mean the draft is NOT ready.

1. Only then proceed to Step 3 (validate.py — the hard gate).

GATE rule: review NEVER blocks the tooling (it leans on a flaky chat API and a re-render is cheap). validate.py is the deterministic hard gate. The override log makes "we saw the finding and chose to ship it" auditable — review.py / validate.py never read it; it is a record for the human in the loop.

Override block shape — work_dir/narration_review_override.md (append-only):

## Override — <date>
- Finding: segment 4 / category=hallucination
- Reviewer said: "‘他早已知情’无画面/对白依据"
- Decision: KEEP — grounded in the --context synopsis (s2 reveal); reviewer lacked that context.
- Signed: <agent/human>

Step 3 — validate

python3 scripts/validate.py --work-dir <work_dir> --mode full   # or --mode cut

Writes narration_lint.json; in full mode rewrites narration.json with quiet-window alignment. Fix any lint errors and re-run until clean.

Cut mode (long video → short recap)

The orchestrated video-recap --edit-mode cut flow is cut-first / narrate-second:

1. Pass 1: write work_dir/clip_plan.json only, using original-time source ranges to keep.
2. The CLI renders edited_source.mp4 and rebuilds the brief with kept clips on the output timeline.
3. Pass 2: write narration.json directly in output time (0..edited_source.mp4 duration). Validate with --mode cut_output via the orchestrator.

The direct video-cut legacy path can still map original-time narration, but the orchestrated recap path does not use that remap.

{"target_duration": "10m", "clips": [{"start": 12.0, "end": 38.0, "reason": "冲突开端"}]}

剪辑模式写作要点（解说要对上剪后的画面，不是原片）：

• Pass 1 只选片段：clip_plan.json 使用原片 source timestamps，挑出完整故事弧，不写解说。
• Pass 2 按实际成片写：edited_source.mp4 已存在时，brief 会列出 kept clips 的 OUTPUT ranges；解说 beat 数量按实际剪后时长估算，不按原片时长。
• 时间轴是 OUTPUT time：narration.json 的 [start,end] 直接落在剪后成片 0..total 时间轴上，不会再做 source→output 映射，也不会静默丢弃。
• 按成片顺序讲：围绕输出时间轴讲一条连续故事线，块尾给原声留白铺垫，下一块承接留白里的对白/动作。

片名/题材明确但缺乏剧情上下文时，先按 背景调研指南 写 background_research.json 再写解说——否则解说只能"看图说话"。brief 在 substrate 偏薄时会把密度目标降为上限而非配额：宁可少写、写实，也不要为凑数堆画面描述。

What this skill does NOT do

• Does NOT run ASR/VLM or analyze the video — it consumes the understanding index.
• Does NOT synthesize audio or render video.
• review.py does NOT edit narration.json and does NOT block the pipeline — it is advisory.
• validate.py does NOT rewrite the meaning of the text — it only checks/aligns timing and quiet windows.