AGENT SKILL

recording-demo-videos

Create or refresh demo videos of a repository's executable applications. Use when asked to record application demos, product walkthroughs, or a video for each app. Discover the applications, run them locally, and deliver verified WebM recordings with edge-tts voice narration, captions, posters, chapters, and rerun commands. Covers web, CLI, API, worker, and native applications; not promotional animation or video editing unrelated to running software.

License 许可
MIT
最近更新
2026-10-19
Stars
★ 1
Forks
0

01 前言 Front Matter

name
recording-demo-videos
description
Create or refresh demo videos of a repository's executable applications. Use when asked to record application demos, product walkthroughs, or a video for each app. Discover the applications, run them locally, and deliver verified WebM recordings with edge-tts voice narration, captions, posters, chapters, and rerun commands. Covers web, CLI, API, worker, and native applications; not promotional animation or video editing unrelated to running software.

02 资源文件 5

03 SKILL.md 正文

Recording demo videos

Create a watchable demonstration of each executable application: real execution, spoken narration, readable captions, and verified outcomes. The default is one continuous successful take at 1280 × 720, roughly two to five minutes per application, with a synthesized voice track. Narration is mandatory: a video without an audible narration track is not a finished demo. Adjust length to useful workflow coverage rather than padding or rushing it. Honor a user's requested applications, format, and presentation overrides.

Resolve supporting references relative to this installed skill folder; resolve application files and docs/demo/ relative to the consuming repository.

1. Understand the repository and identify applications

Read applicable agent instructions, README files, workspace/project manifests, entrypoints, startup scripts, tests, and existing demo tooling. Read useful content under docs/specs/ and docs/detailed-designs/ when present. Neither directory is required: infer behavior from source, tests, and runtime inspection when absent. If documents disagree with the implementation, demonstrate what actually works and record the discrepancy; do not claim unimplemented requirements.

Inventory every executable application before selecting scenes. Record its purpose, audience, entrypoint, run command, local endpoint or interaction surface, dependencies, authentication, and candidate workflows. Count independently runnable first-party APIs and workers even when a web app also depends on them. Separate user-facing apps served by one host still deserve separate videos. Exclude libraries, migrations, test runners, repository maintenance scripts, and third-party infrastructure unless the user explicitly wants to demonstrate those tools.

Assign stable lowercase hyphenated slugs, preserving established demo names. Disambiguate duplicate names with their subsystem. Account for every application as recorded, blocked with a reason, or excluded by the user's scope. A repository with no executable applications needs an explanation, not an invented demo.

2. Choose workflows and the recording method

Build a short storyboard per application: purpose, starting state, actions, observable assertions, narration, and a useful final state. Write the narration as short spoken paragraphs, one per opening card, chapter, or step, in second person and plain language; the same text becomes the on-screen captions. Prefer complete user journeys over touring every route or endpoint. Include meaningful validation or permission behavior when it explains the product. Avoid exposing credentials, private records, tokens, or sensitive shell history in footage and diagnostics.

Reuse existing acceptance helpers and demo conventions when suitable. Select the capture method from available tools and the application's real interaction surface:

If apps share data, order their stories by dependency and reuse deterministic synthetic fixtures. Explain which earlier recording establishes later state. Provide a command that reproduces the whole sequence from a clean environment; an individual rerun must prepare its prerequisites or name the required sequence.

3. Prepare a reproducible local run

Verify the required runtimes, locked dependencies, capture tools, database, configuration, and local credentials, plus the narration tools: Python with the edge-tts package, internet access for synthesis, and ffmpeg with ffprobe, libopus, and libvpx. Prefer repository setup/build scripts. Use normal authorized local setup to resolve missing dependencies; if credentials, platform support, or external access remain unavailable, report exactly what is needed and continue independent applications.

Use disposable demo storage and synthetic identities. Before any reset or cleanup, verify the resolved database, directory, or container belongs to this run. Never inherit a reset-on-start setting against an existing development database. Do not reuse a listening server unless its identity, configuration, and data isolation are established; otherwise choose unused ports. Bind demo-only services locally.

Use real application services and persistence where the application requires them. Existing local adapters for mail, payments, AI, or other external integrations are acceptable when disclosed. Do not substitute mocked product responses or remove authentication to make the demo pass. Avoid actual external sends or purchases without authorization already covering them.

Create application-specific recording scripts in the repository's existing test or script structure. Keep recordings separate from the ordinary acceptance run and its commit gates. Give startup, individual actions, the whole take, and teardown finite timeouts. Check readiness before recording; do not fill the video with build logs or server startup waits.

Track owned processes and resources from their creation. Enclose startup as well as recording in cleanup handling so partial setup failures also clean up. On Windows, launch background helpers hidden. Stop only owned processes, restore changed environment settings, and report any cleanup failures and resources left. Keep temporary video, trace, credentials, and logs in an ignored run directory.

4. Synthesize the narration

Synthesize the voice track with the free edge-tts Python package (Microsoft Edge online read-aloud; no API key, Azure subscription, or paid Speech service). Install it with python -m pip install edge-tts using the interpreter named by PYTHON, and check connectivity with python -m edge_tts --list-voices. Use the same voices as the creating-narrated-videos skill: en-US-AndrewMultilingualNeural for the narrator and en-US-AvaMultilingualNeural for a second speaker such as a labelled customer or administrator; EDGE_VOICE and EDGE_VOICE_2 override them. Use one voice set across every application in the same demo index.

Keep the narration text in a per-application <app-slug>-narration.md whose paragraphs appear in capture order, each tagged with the chapter or step it belongs to. Edge does not support custom SSML, so keep acronyms and code identifiers in a pronunciations.json lexicon of plain-text spoken replacements, apply it before synthesis, and spot-check difficult entries. Captions keep the written form; the voice track uses the spoken form.

Synthesize before recording, one clip per paragraph, into the ignored run directory. Normalize the clips to a common format, measure each with ffprobe, and make every presentation hold at least as long as its clip plus a short margin so the narrator finishes before the next action. During capture, log the timestamp at which each paragraph's caption appears. After the video finalizes, place each clip at its logged time, corrected by the same offset used for the chapter times, pad the gaps with silence, loudness-normalize the result, and mux it into the WebM as an Opus track with ffmpeg without re-encoding the video. The video stream, its duration, and the chapter times must not change.

If Python, edge-tts, internet access, or ffmpeg is unavailable, that is an actionable blocker for narration: keep the recording and its narration text in staging, state exactly which commands remain, and do not promote or describe a silent video as a completed demo. Never substitute silence, a text transcript, or synthetic placeholder audio for the narration.

5. Record and verify

Use readable pacing, a brief opening, chapter transitions, and captions explaining the user's outcome. Keep captions clear of the controls or output being shown. Presentation delays are for listening and reading; readiness must use observable conditions. Assert actual results before narrating that they succeeded: persisted state, calculated values, loaded media, returned responses, or completed jobs.

Record one continuous successful take per application. For browser capture, keep the main story in one recorded page. Multiple actors may switch authenticated sessions within that story without exposing secrets. Do not splice failed scenes into a successful recording. Disable automatic retries; diagnose a failed take and re-record it in full after a justified harness or setup correction.

Fix local setup and recording-harness problems within the request. Product bugs are outside this skill's default repair scope: capture the failing step and expected versus actual behavior, mark the affected demo blocked, and finish independent apps. Do not repeatedly retry unchanged failures or weaken assertions. User authorization for product repairs can expand that scope explicitly.

After capture has finalized and the narration is muxed, open and inspect the encoded WebM itself. Check playback, measured duration and dimensions, beginning and ending, each chapter, caption readability, and important visual outcomes. Confirm with ffprobe that the file has one video stream and one Opus audio stream of the same duration, then play the start of each chapter and verify the spoken paragraph matches its caption and begins within about a second of the action it describes. Sample frames are useful but do not replace playback review for pacing. Verify media really loaded; an image's URL or alt text alone is not evidence. Derive chapter times from the recording timeline, verify them against playback, and choose a representative poster from the successful footage. A nonempty file is not sufficient validation.

6. Deliver artifacts and rerun instructions

Stage each application's video, poster, narration text, and chapter metadata outside final paths. Promote the verified set only after its assertions and video review pass. Back up existing files and restore them if promotion fails; do not mix an old video with new chapter times or a new poster. Preserve unrelated demos and manual README content. If a rerun fails, distinguish any retained prior video from this run.

Default deliverables in the consuming repository:

docs/demo/
├── README.md
├── pronunciations.json
├── <app-slug>.webm
├── <app-slug>-poster.png
└── <app-slug>-narration.md

Repeat the video, poster, and narration files for every application. Synthesized clips and the assembled audio track stay in the ignored run directory; the narration text and lexicon are the reproducible sources. The README must include:

  • An application inventory with links, purpose, and recorded/blocked status.
  • Measured video duration, dimensions, file size, audio codec, narrator voices, and linked poster and narration text per recording.
  • Chapter timestamps and the verified workflows they demonstrate.
  • Exact setup and rerun commands, working directories, tool prerequisites, configuration variable names, data preparation, and cleanup behavior.
  • Recording order, relevant source/test links and revision when available, integration substitutions, product limitations, and actionable blockers.

Record the actual commands, not an assumed universal npm run demo. Keep secrets out of documentation. Include reproducible recording sources, but keep transient logs and failure artifacts out of the deliverable set. Report completion only when every in-scope application has a reviewed video; otherwise report partial completion with explicit blockers and links to the successful artifacts.

04 同仓库更多 Skills

QuinntyneBrown
QuinntyneBrown/quinntyne-brown-studio

writing-screen-composition

Create or update docs/screen-composition.md, a document that itemizes every page, dialog, notification and shell part from the HTML mocks and lists, for each, the component-library components it shall be composed of plus the UI that must be built locally. Also ensures AGENTS.md has the "Components library is mandatory (CRITICAL)" section. Use when the user mentions screen composition, which components a page uses, mapping mocks to a component library, local vs library UI, or before implementing an application against a component library; and whenever docs/mocks and a component catalogue exist but docs/screen-composition.md does not or is out of date.

★ 1
QuinntyneBrown
QuinntyneBrown/quinntyne-brown-studio

writing-requirements

Create and manage software requirements documents (L1 high-level and L2 detailed with acceptance criteria) in docs/specs/. Use this skill whenever the user mentions requirements, specs, specifications, user stories, features to build, acceptance criteria, ATDD, or wants to define what a system should do before coding. Also use when the user says things like 'what should we build', 'let's plan the feature', 'write up what this needs to do', 'define the behavior', or describes functionality they want implemented -- even if they don't explicitly say 'requirements'. If the user is about to start coding a feature and there are no requirements yet, proactively suggest using this skill first. Requirements come before code. Always.

★ 1
QuinntyneBrown
QuinntyneBrown/quinntyne-brown-studio

writing-html-mocks

Write static HTML mocks in docs/mocks/ for EVERY page, dialog and notification of a product in EVERY state (default, loading, empty, error, invalid, busy, success, no-results, forbidden, stacked...), responsive, light and dark, keyboard-accessible, with real copy and a coverage matrix. Use this skill BEFORE writing any frontend code, component, template, view or page; whenever the user says mock, mockup, wireframe, prototype, UI, UX, screen, layout, dialog, modal, toast, banner, empty/error/loading state, "what should this look like", "show me the UI", or describes a feature with a visible result; and whenever docs/specs or docs/detailed-designs exist but docs/mocks does not. Also use it to add missing states to existing mocks. Mocks come before implementation: if asked for frontend code with no mocks, say so and offer this first. Pair with extracting-design-systems to turn finished mocks into docs/design-system/.

★ 1
QuinntyneBrown
QuinntyneBrown/quinntyne-brown-studio

writing-feature-prompts

Write a ready-to-run implementation prompt for a feature request or a bug fix in the current repository, save it outside the repository and open it in a text editor (Notepad on Windows). A feature prompt drives mocks, design system, requirements, detailed design, estimate update and incremental acceptance-test-driven implementation end to end without stopping; a bug-fix prompt drives reproduce, root cause, fix and documentation alignment. Use whenever the user asks to "create a prompt", "write a prompt for this feature" or "for this bug", wants a "prompt to build X" or "prompt to fix Y", or asks to open a prompt in Notepad.

★ 1
QuinntyneBrown
QuinntyneBrown/quinntyne-brown-studio

writing-design-documents

Create software design documents and architecture write-ups from existing L1/L2 requirements in docs/specs/. Use for detailed designs organized by subsystem and feature, with requirement traceability and rendered PlantUML C4, class, and sequence diagrams in docs/detailed-designs/.

★ 1
QuinntyneBrown
QuinntyneBrown/quinntyne-brown-studio

writing-agent-instructions

Writes the agent instruction files a repository needs so coding agents know how to work in it - AGENTS.md plus the CLAUDE.md, GEMINI.md, and .github/copilot-instructions.md pointers - generated from a description of what the project shall be, for a project that does not exist yet. Use this whenever someone mentions AGENTS.md, CLAUDE.md, GEMINI.md, copilot-instructions, "agent instructions", "agent context files", "coding agent guidance", or asks to set up, scaffold, bootstrap, or initialise the conventions for a new project or repository. Use it too when someone describes a project they are about to start and wants an agent to follow its conventions, even if they never name any of these files - if they are standing at an empty directory describing what they intend to build, this is the skill.

★ 1