/style concise — and every reply from now on is terse. /style step-by-step — and the model narrates numbered steps. /style off — back to the project default. One command per session, persisted across restarts, zero changes to the agent loop.

✨ Features

🚀 Quick start

# 1. Install — the package is a bundle layer, so one command composes
#    storage + storage-json + storage-domain + the plugin row:
dsh plugin --profile <name> add dsh-output-styles

# 2. Boot and switch
dsh --profile <name>
/style               # → output style off, then one line per style
/style concise       # → switched to concise
/style Diagrams first  # → names with spaces work too
/style off           # → back to the project default

The layer is idempotent over web profiles (insert-by-id replaces same-id rows), which compose storage in box. For the Web picker, add the client row to the profile:

- id: output-styles-client
  name: 'dsh-output-styles/client'

🎬 Demo

You > /style
      output style off
      concise — Terse, direct answers — minimal prose, no preamble. (Daily coding work, tool-heavy sessions, or when prompt length matters.)
      explanatory — Educational answers with short "Insights" that teach as you work. (Learning a codebase, onboarding, …)
      formal — Formal, precise prose with complete sentences and defined terms. (Reports, documentation, release notes, …)
      learning — Collaborative learn-by-doing mode with short "Insights" and small hands-on steps for the user. (Pairing, onboarding, …)
      proactive — Execute immediately, assume reasonable defaults, and prefer action over planning. (Routine multi-step work, …)
      step-by-step — Numbered reasoning steps with explicit intermediate results. (Debugging, design decisions, …)

You > /style concise
      switched to concise

You > 请只用一句话介绍你自己。
AI  > 我是运行在 DeepSeek Harness 插件化平台上、基于 deepseek-v4-pro 模型的 AI 编码代理。

🧠 How it works

flowchart LR
    U[You type /style concise] --> C[command registry]
    C -->|command/run logged| L[(session log)]
    C -->|put {style, source}| D[(output_style domain)]
    D --> R[OutputStyleRuntime]
    R -->|body at every assembly| S[systemPrompt section order 90]
    S --> M[Model request]
    M -->|full system prompt| H[request/header logged]

Everything the model sees is reconstructable from the session log — no new session event type, no agent-loop changes. The style name comes from command/run, the exact injected text from request/header, and the provenance marker { kind: 'plugin', plugin: 'dsh-output-styles' } rides in the domain record. Styles apply to the main conversation only; subagent sessions keep their own prompts (matching Claude Code).

⚙️ Configuration

Every tunable is a validated Schemastery Config field (invalid values fail the load):

📚 Style library

---
name: concise
description: Terse, direct answers — minimal prose, no preamble.
whenToUse: Daily coding work, tool-heavy sessions, or when prompt length matters.
keep-coding-instructions: true
---

You are in the concise output style for this conversation.
- Lead with the direct answer; skip preamble, restatements, and filler.
- 回答语言跟随用户语言：中文提问用中文回答，英文提问用英文回答。

Frontmatter fields:

{ "name": "explain", "description": "Explain like a teacher.", "prompt": "Teach in small steps." }

⌨️ Command reference

🎨 Renderer protocol

The output.render.* protocol turns the presentation layer into an extension point. A renderer is a pure presenter — presenter(text, context) maps args to display data, never touches the DOM — matched by tool name and content type, ordered by priority:

// Third-party plugin registering a custom renderer (register() returns the disposer)
ctx.effect(() => ctx.outputRenderers.register({
  id: 'sql-table',
  name: 'SQL table compactor',
  description: 'Truncates oversized SQL result sets to the head plus a row count.',
  match: [{ tool: 'sql', contentType: 'text' }],
  priority: 20,
  presenter: (text, context) => compactRows(text, 50),
}))

• Waterfall first: every render request passes through output.render/before ({ text, context }) — listeners transform the request and must call next(); returning without it short-circuits the pipeline.
• Rules: rules: [{ match: { tool: 'bash' }, style: 'concise' }] names the renderer for matching requests (tool, content type, or exact session); ties break by priority, then rule order. Rules live in cordis.yml and the output-style-rules settings section.
• Built-ins: concise (whitespace compaction + budget truncation) and step-by-step (consistent step numbering) — ids mirror the two headline style names.
• Auditability: every render result carries { original, rendered, rendererId, changed }. The rendered text is what surfaces; the original text is the session log itself, and the render application is deterministic — so rendered output and its source always reconstruct together.
• Full protocol reference (including a worked third-party example): docs/renderer-protocol.md (中文: docs/renderer-protocol.zh.md).

🖱️ Web picker

The dsh.client entry decorates the host /style command's bare invocation with a popup picker: an "off" row plus one row per library style (description · whenToUse), the active row marked. Picking submits /style <name> through the command Remote, so every switch keeps the host's durable command lifecycle and the style projection stays the single displayed fact. The picker follows the Web UI's shipped zh/en locale pair.

🔍 Conflict check

Screened against the DSH ecosystem before development (2026-08 snapshot): no style/output-style repository under topic:dsh-plugin, no output-style category in the four major awesome lists, and no entry in the dsh-hub catalog. The closest neighbors — dsh-soul-md (persona) and dsh-claude-marketplace (output styles explicitly deferred to v0.2+) — are adjacent, not conflicting.

🆚 Differences from Claude Code

🧪 Development

pnpm install
pnpm run typecheck   # both tsc projects
pnpm test            # vitest — 107 tests
pnpm run verify      # typecheck + tests + self-contained (the prepublishOnly gate)
pnpm run build       # lib/ artifacts (host + client bundles)
pnpm pack            # tarball for dsh plugin add

Releases: pushing a v* tag whose suffix matches the package.json version triggers the Publish workflow — full verification, then an npm publish with provenance. Any npm publish also passes the verify gate through prepublishOnly.

Structure follows the omdsh-dev/plugin-template: src/index.ts (plugin metadata), src/config.ts (schema), src/runtime.ts (runtime service + activation), src/invariant.ts (invariants), src/client/ (Web picker), styles/ (built-ins).

👥 Contributors

Thanks to everyone who helped build this project:

• @PerryLink — author and maintainer: plugin architecture, style library, bundle install, Web picker, five-language docs, and CI/release tooling.

Found a bug or an idea? Open an issue or send a pull request — contributions in any language are welcome.

PerryLink DSH Plugin Family

This project is one of the 15 DeepSeek Harness plugins maintained by PerryLink. If this one helps you, the others likely will too:

📄 License

Apache-2.0 © 2026 dsh-output-styles contributors

_{Topics: dsh · dsh-plugin · deepseek-harness · output-styles · claude-code}