Lumos-789 · Nov 13, 2026
@@ -0,0 +1,33 @@
+{
+  "name": "delivery-video",
+  "version": "1.2.2",
+  "description": "Turn a feature that actually runs into a trustworthy delivery demo video, or a defect into a traceable feedback video: real screen recordings, edge-tts narration, word-synced subtitles and a deterministic quality gate.",
+  "author": {
+    "name": "Lumos-789",
+    "url": "https://github.com/Lumos-789"
+  },
+  "homepage": "https://github.com/Lumos-789/delivery-video-plugin",
+  "repository": "https://github.com/Lumos-789/delivery-video-plugin",
+  "license": "MIT",
+  "keywords": [
+    "delivery",
+    "demo-video",
+    "bug-report",
+    "screen-recording",
+    "quality-gate",
+    "pi",
+    "dsh"
+  ],
+  "skills": "./skills/",
+  "interface": {
+    "displayName": "Delivery Video",
+    "shortDescription": "Record and gate delivery demo videos",
+    "longDescription": "Records real screen sessions of working software, renders narration and word-synced subtitles, and runs a deterministic quality gate before the video is delivered. The same pipeline produces defect/feedback videos with severity and repro/actual chips plus a hashed issue report. No static screenshots, mocks or stale footage.",
+    "developerName": "Lumos-789",
+    "category": "Productivity",
+    "capabilities": [
+      "Read",
+      "Write with confirmation"
+    ]
+  }
+}
@@ -0,0 +1,27 @@
+{
+  "name": "delivery-video",
+  "version": "1.2.2",
+  "description": "Record, render and gate trustworthy delivery demo videos — or defect/feedback report videos — of real, working software: real screen recordings, edge-tts narration, word-synced subtitles and a deterministic quality gate.",
+  "description_i18n": {
+    "en": "Record, render and gate trustworthy delivery demo videos — or defect/feedback report videos — of real, working software: real screen recordings, edge-tts narration, word-synced subtitles and a deterministic quality gate.",
+    "zh-CN": "把真实运行的功能录成可交付的演示视频，或把缺陷录成可追溯的问题反馈视频：真实录屏、edge-tts 旁白、逐字对齐字幕与确定性质量闸，产出 1920x1080 成片与完整证据链。"
+  },
+  "author": {
+    "name": "Lumos-789",
+    "url": "https://github.com/Lumos-789"
+  },
+  "homepage": "https://github.com/Lumos-789/delivery-video-plugin",
+  "repository": "https://github.com/Lumos-789/delivery-video-plugin",
+  "license": "MIT",
+  "keywords": [
+    "delivery",
+    "demo-video",
+    "bug-report",
+    "screen-recording",
+    "acceptance",
+    "quality-gate",
+    "edge-tts",
+    "ffmpeg"
+  ],
+  "skills": "skills"
+}
@@ -0,0 +1,88 @@
+# Changelog
+
+## 1.2.2
+
+- **Fixed the dead black-frame check.** `parse_detects` required a space after the
+  colon, but ffmpeg prints blackdetect as `black_start:6 black_end:11.966667` with
+  no space (silencedetect/freezedetect do print one). A 12s video whose last 6s are
+  pure black passed the gate with empty `black_spans`; the regex now accepts both
+  formats, so `B101` fires again.
+- Detect spans that run to the end of the video are no longer dropped:
+  freezedetect emits only `freeze_start` (no `freeze_end`) when the freeze reaches
+  EOF, and such spans are now closed at the total duration, so a frozen tail
+  triggers `T301`/`T304`.
+- Feedback tamper-proofing can no longer be switched off by editing `episode.json`
+  (which sits outside every hash chain): `verify_delivery.py` now checks the
+  report/markdown hash bindings whenever the gate report carries a `feedback`
+  block, not only while episode.json still declares `intent: "feedback"`.
+- `preflight.py` no longer demands a `size` field or a 1920x1080 resolution from
+  demo-stage evidence, matching the documented demo-stage contract
+  (sha256/fps/frames/duration) and `verify_delivery.py`. The build-name rejection
+  message now attributes the check to preflight — render.py accepts any directory
+  name; preflight is what rejects it.
+
+## 1.2.1
+
+- Docs fix: `references/recording.md` still claimed "the renderer crops to a 1080p
+  canvas" — the opposite of the 1.2.0 fix. It now says recordings are scaled/padded
+  onto the 1080p canvas and the full frame is shown unless a segment declares an
+  explicit `crop`, matching SKILL.md and `episode-and-review.md`.
+- `render.py` and `preflight.py` no longer duplicate the crop validation: one shared
+  `crop_error()` (in `render.py`) is collected by preflight and raised by render.
+  Both entry points reject exactly the same inputs as before; render's message for a
+  negative-origin or non-positive crop now names that problem instead of the old
+  "exceeds source".
+- Comment cleanup: dropped the `LAYOUTS` block comment duplicated by
+  `segment_filter`'s docstring, the behavior-restating half of the polyphone comment,
+  and a divider comment.
+
+## 1.2.0
+
+- **Root-caused the "only the middle of the screen is recorded" failure.** `render.py`
+  used to apply a hard-coded `DEFAULT_CROP = [260, 64, 1340, 780]` to every segment
+  that did not declare `crop`, silently trimming 260 px off the left and 320 px off the
+  right of any full-window 1920x1080 recording (and `preflight.py` validated browser
+  takes against the same silent default while demo-stage takes were validated as full
+  frame — the two defaults even disagreed). Frame geometry is now opt-in everywhere:
+  `crop` absent = the whole recording is shown; an explicit crop is bounds-checked
+  against the real source pixels in both preflight and render and fails loud.
+- New episode-level `layout` field: `banded` (default, unchanged dark-canvas look with
+  the 1800x866 content box) or `fullbleed` (recording scaled onto the full 1920x1080
+  canvas; 16:9 sources map 1:1; case chip stays top-left and subtitles get a dark
+  backing box so they stay readable over light pages). Use `fullbleed` when every
+  pixel of the screen must stay visible.
+- Polyphone lint in `preflight.py`: narration containing 行/重/还/都/长/得/着/差/数/干/应/结/系/弹/屏
+  produces a per-occurrence warning with context (e.g. table-row 行 háng vs walking 行
+  xíng — prefer 一条/这一单), so TTS mispronunciations are caught before render instead
+  of after delivery.
+- Tests: `tests/test_render.py` covers full-frame default, explicit-crop placement and
+  fullbleed geometry.
+
+## 1.1.2
+
+- Feedback reports are now fully bound to the evidence chain. `render.py` writes the
+  artifact SHA256 plus, into `quality_gate/report.json`, a normalized hash of the
+  `feedback.json` fields and the exact `feedback.generated.md` bytes.
+- `verify_delivery.py` recomputes both hashes. Editing `issue.title`/`issue.actual` (or
+  any other report field), editing the markdown, deleting the markdown, or rendering
+  with an older report that has no feedback binding now fails with a specific message
+  instead of `DELIVERY=PASS`. The video-side binding (artifact SHA256, truncated video,
+  stripped audio, re-encoded resolution) is unchanged.
+- Tests: `tests/test_feedback_binding.py` covers clean pass, edited title, edited actual,
+  edited markdown and deleted markdown; `tests/test_render.py` covers canonical hash
+  normalization. Feedback artifacts rendered by 1.1.1 must be re-rendered to gain the
+  binding.
+
+## 1.1.1
+
+- Feedback/defect report mode: `intent: "feedback"` plus an `issue` block adds severity
+  and phase chips, a repro-steps overlay, and writes `feedback.json` and
+  `feedback.generated.md` bound to the artifact SHA256. Quality-gate rules are unchanged.
+
+## 1.0.x (2026-10-31 → 2026-11-02)
+
+- 1.0.0 — first ZCode release: real-recording pipeline (`preflight`/`render`/`verify`), the built-in `delivery-v1` gate, bilingual READMEs.
+- 1.0.1 — narration/TTS, subtitle line-breaking and low-bitrate false-positive fixes; preflight checks render dependencies early.
+- 1.0.2 — default warnings cleaned up; auditable notices kept.
+- 1.0.3 / 1.0.4 — DSH runtime skill registration, then registration scoped to each Agent (plugin shell: `package.json` + `index.js` + `cordis.patch.yml`).
+- 1.0.5 / 1.0.6 — desktop-environment hardening; case-driven recording guidance.
@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Lumos-789
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
@@ -0,0 +1,76 @@
+# delivery-video
+
+A skill plugin that turns a feature that **actually runs** into a trustworthy delivery demo video — real screen recordings, synthesized narration, word-synced subtitles, and a deterministic quality gate that refuses to pass a fake or broken demo.
+
+Most "AI demo videos" are stitched from static screenshots, mock pages or stale recordings. This plugin enforces the opposite discipline: every segment must point at a real recording with a traceable time range, the last shot must end on a verified result state, and the artifact must pass hard checks for resolution/FPS/codec, audio presence, loudness, black frames and freeze frames before anything is posted. Bitrate density is reported as an informational notice; text still needs a 1:1 visual check. The uploaded file is always the gated file, bound by SHA256.
+
+In DeepSeek Harness, the skill is registered in both the Host and each live Agent scope. That lets the installed plugin win over an older same-name copy in `~/.agents/skills`, even when an Agent preset mounts the filesystem skill provider in a nearer layer.
+
+## What it does
+
+1. **Verify first, record second** — the workflow requires evidence that the current version really runs (git SHA, tickets, request IDs) before any recording.
+2. **Case-driven structure (recommended)** — for multi-case deliveries the skill suggests listing the acceptance cases first, recording one real run per case, then assembling segments in execution order with a result shot per case. Optional, not a gate.
+3. **episode.json recipe** — each segment binds one real take + time range + one narration line + optional crop/case chip.
+   Set `intent: "feedback"` with an `issue` block to produce a defect report instead of a
+   delivery: severity and phase chips, a repro-steps overlay, and both `feedback.json` and a
+   ticket-ready `feedback.generated.md`. The gate report binds the artifact SHA256, a
+   normalized hash of the `feedback.json` fields and the exact markdown bytes, so
+   `verify_delivery.py` fails when the report was edited after rendering, or the markdown is
+   missing or was edited after rendering. Copy `examples/feedback-episode.example.json` as a
+   starting point. The same quality gate applies.
+4. **Self-contained renderer** — `render.py` synthesizes narration via edge-tts (word-boundary clocks), times each segment to its narration (0.3x–3x speed is hard-checked, never clamped — out of range fails), draws the label chip and word-synced subtitles, concatenates, loudness-normalizes to −14 LUFS, and runs the built-in quality gate. Output: `screen_demo.mp4` (1920×1080, 30fps, h264 with AAC audio — a non-AAC track is a gate warning `E102`, not a failure), `timeline.json`, cover/preview frames, `quality_gate/report.json`.
+5. **Fail-loud checks** — `preflight.py` validates the recipe before rendering (structure, intervals, crops, browser-recording evidence, TTS cache contracts, narration red lines); `verify_delivery.py` verifies the finished artifact, gate binding and optional independent visual review afterwards.
+6. **Attachment gate** — `render.py gate` runs the same hard checks on any mp4 before it goes into a ticket, chat or customer channel.
+
+## Prerequisites
+
+- `python3` 3.9+
+- `ffmpeg` and `ffprobe` on PATH
+- `pip install "edge-tts>=7.3.1" Pillow`
+
+`edge-tts` 7.2.8 and older can return HTTP 403 from the upstream endpoint. Use 7.3.1 or newer; the renderer fails early with an upgrade command when the installed version is too old.
+Run `preflight.py`, `render.py` and `verify_delivery.py` with the same Python interpreter. Preflight fails early when ffmpeg/ffprobe, Pillow, a supported edge-tts version or a CJK font is missing.
+
+Platform setup:
+
+- macOS: `brew install ffmpeg`, then `python3 -m pip install "edge-tts>=7.3.1" Pillow`
+- Windows: install FFmpeg with `winget install --id Gyan.FFmpeg -e` (or `choco install ffmpeg`). If `python3` is not available, use `py -3` consistently and run `py -3 -m pip install "edge-tts>=7.3.1" Pillow`
+- Linux: install `ffmpeg` and a CJK font such as `fonts-noto-cjk` with your package manager, then install the Python packages
+
+The renderer needs a CJK font for Chinese labels and subtitles. macOS and Windows normally provide one; minimal Linux images often do not, and preflight reports the missing font before rendering.
+
+## Side effects (disclosed for review)
+
+- **Network**: narration synthesis calls the public edge-tts (Microsoft Cognitive Services) endpoint; no other network access.
+- **Files**: writes only inside the build directory you pass in (`tts/` cache, per-segment clips, final mp4, gate reports).
+- **Commands**: runs `ffmpeg`/`ffprobe` locally; the skill never uploads, publishes or deletes anything by itself.
+
+## Quick start
+
+Install: `dsh plugin add /path/to/delivery-video-plugin` (DSH bundle). It exposes the skill in `skills/delivery-video/`.
+
+1. Record real takes of the feature with your usual tool (macOS `Cmd+Shift+5`, OBS, or a scripted browser screencast — see the skill's recording reference). For two or more acceptance cases, list them first using `examples/cases.example.md` as a template, then record one real run per case.
+2. Create `build/screen_demo-<slug>/episode.json` (copy `examples/episode.example.json`).
+3. Run the three scripts (use `py -3` instead of `python3` on Windows when needed):
+
+```bash
+python3 <skill>/scripts/preflight.py build/screen_demo-myfeature
+python3 <skill>/scripts/render.py render build/screen_demo-myfeature
+python3 <skill>/scripts/verify_delivery.py build/screen_demo-myfeature
+```
+
+Run the unit tests with:
+
+```bash
+python3 -m unittest discover -s tests -v
+```
+
+Suggested trial prompt after install: *"用 delivery-video 把这个项目刚完成的功能录成一段 90 秒交付演示视频"* — the skill will walk verification, storyboard, recording guidance, recipe assembly, render and gating.
+
+## Same-name skill copies
+
+ZCode 0.17.x can keep a user-level skill and a plugin skill with the same name. If an older `~/.agents/skills/delivery-video` still exists, the unqualified `/skill delivery-video` command may select that copy. Select `delivery-video:delivery-video` in the composer, or use `/skill delivery-video:delivery-video <task>`, to run the installed plugin.
+
+## License
+
+MIT © Lumos-789