Read product-facing context
Use README files, public docs, changelogs, package metadata, landing-page copy, and selected screenshots or clips.
Use the repository context already in your DSH workspace to shape three Angles launch-video concepts, choose the strongest story, and render only after confirmation.
Pinned install
dsh plugin --profile demo add github:anglesvideo/anglesvideo-dsh-plugin#d0527be50716a208b57a0d1ddbc64a846ace1d31No raw repository upload · Three concepts first · Explicit render confirmation
Context boundary
read README.md
read docs/public
read CHANGELOG.md
skip .env
skip customer-data

Agent response
Workflow
DeepSeek Harness is a strong fit because launch-video creation is not a single prompt. It is a sequence of visible decisions.
Use README files, public docs, changelogs, package metadata, landing-page copy, and selected screenshots or clips.
Convert repository context into audience, pain point, promise, proof, use case, and launch constraints.
Compare distinct hooks, narratives, scene plans, media requirements, and recommended Angles templates.
Render only after a human chooses the concept and explicitly confirms that Angles should create the MP4.
Prompt shape
The first DSH workflow should make the agent produce reviewable launch directions before it calls any render endpoint.
Create a 30-second product launch video for this repository.
Read only product-facing context: README, public docs, changelog,
package metadata, landing-page copy, and screenshots I provide.
Before rendering, show me three distinct launch angles with target viewer,
hook, narrative, recommended template, media needs, and claims to verify.
Do not render until I choose an angle and explicitly confirm.Safety boundary
A useful launch video usually comes from README, docs, screenshots, changelog, and public positioning. Sensitive implementation data should stay out of the workflow.
Do not upload raw repositories to Angles.
Skip .env files, credentials, private keys, database exports, customer data, and production logs.
Send a structured product summary instead of unrelated implementation files.
Preview missing media or claim risks before spending a render allowance.
Status
DeepSeek Harness is public, active, and plugin-native, but its own documentation marks the project as a developer preview. The Angles integration should ship with tested versions and pinned compatibility notes.
DeepSeek describes Harness as a plugin-composable agent runtime where capabilities can be swapped or recomposed.
Open sourceCommunity projects are currently discoverable through the GitHub dsh-plugin topic.
View topicThe Angles DSH bundle is public with a real repository, tested tool boundary, and pinned-commit installation path. Expect updates while DeepSeek Harness remains in developer preview.
View the Angles pluginFAQ
Yes. The public bundle is https://github.com/anglesvideo/anglesvideo-dsh-plugin. It is an early integration for the DeepSeek Harness developer preview, so install a pinned commit and expect compatibility updates.
DeepSeek Harness is plugin-native but its official documentation warns that compatibility-breaking changes may happen. The bundle is public and installable, but its README pins the tested DSH/runtime assumptions.
The intended workflow sends a structured product brief, not a raw repository upload. Product-facing context is enough for launch-video angles, and sensitive files should stay local.
Yes. The Angles developer API already supports concept generation, templates, render preview, asset upload, render confirmation, and status polling.
Concept generation, template metadata, render preview, asset upload, confirmed render, and status polling are already available to authenticated developer workflows.