Engineering workflow agent preset for dsh: five gated phases (requirements clarification, plan approval, TDD, parallel subagents, verified finishing) with six workflow skills adapted from obra/superpowers.
Install
# from a prebuilt release tarball
dsh plugin --profile web add "https://github.com/82c86b8z86-stack/dsh-engineering-workflow/releases/latest/download/dsh-engineering-workflow.tgz"
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:82c86b8z86-stack/dsh-engineering-workflow
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
An engineering workflow layer for DeepSeek Harness (dsh). One install adds the 工程工作流 (Engineering Workflow) agent preset — a disciplined-engineer mode with five hard-gated phases — plus six workflow skills that carry the methodology.
The workflow methodology is adapted from obra/superpowers (MIT): brainstorming, writing-plans, TDD, subagent-driven development, and verification-before-completion, reworked for dsh's native tools (plan mode + exit_plan_mode, background subagent/subagent_fork, workflow orchestration, goals, and the preset/skill system).
The five phases
| Phase | Skill | Gate | dsh mechanism |
|---|---|---|---|
| ① Requirements clarification | workflow-requirements |
Intent approved before any code | ask_user_question, one question at a time |
| ② Plan approval | workflow-planning |
Plan approved via exit_plan_mode |
plan mode, todo_write after approval |
| ③ TDD implementation | workflow-tdd |
Failing test before production code | pwsh/bash test runs |
| ④ Parallel subagent execution | workflow-subagents |
Per-task review + ledger | background subagent, send_message, list_agents |
| ⑤ Verified finishing | workflow-verification |
Fresh evidence before claims | full suite + branch-finish menu |
The master skill engineering-workflow routes every non-trivial task to the right phase and enforces the discipline rules (rationalization red flags included).
Install
dsh plugin --profile <name> add github:82c86b8z86-stack/dsh-engineering-workflow
(Or npm install the package into your profile and add dsh-engineering-workflow to dsh.profile.bundles.)
Restart dsh once so the host plugin mounts. On startup it syncs the preset into ~/.dsh/.agent-presets/engineering-workflow; the preset then appears in the new-session preset picker as 工程工作流. The sync is idempotent — upgrading the plugin updates the preset and its skills automatically.
Manual / development fallback without a restart:
node scripts/sync-presets.mjs
dsh re-discovers presets on every roster read, so the synced preset is selectable immediately.
How it works
dsh-engineering-workflow (bundle)
├── cordis.patch.yml inserts one host plugin row
└── lib/index.js host plugin: syncs presets/ → ~/.dsh/.agent-presets,
│ announces the workflow via a system-prompt section
└── presets/engineering-workflow/
├── agent.cordis.yml full toolset composition (adapted from the shipped
│ cordis preset, MIT): shell, filesystem, jobs, goals,
│ plan mode, compaction, delegation (subagent/subagent_fork/
│ workflow/ralph), ask-user, todo, web, skills
├── preset.yml roster metadata (name / description / order)
├── skills/ 6 workflow skills (one SKILL.md per directory)
└── NOTICE attribution
The preset wires its skills through @deepseek-ai/dsh-skill-filesystem with customSkillDirs rooted at the preset's own directory — the same pattern the shipped cordis preset uses, so the skill catalog travels with the preset wherever it is installed.
Verify the install
node scripts/verify-install.mjs
# ✓ engineering-workflow: current (byte-identical)
This compares the bundled preset against ~/.dsh/.agent-presets without writing anything. If it reports stale or missing, run node scripts/sync-presets.mjs (or restart dsh once so the host plugin syncs on mount).
Where the plugin takes effect:
- The preset picker (new-session dialog) lists 工程工作流 — dsh re-discovers presets on every roster read, so a synced preset appears without a restart. Pick it for a new session; existing sessions and the default preset are never switched.
- The system-prompt announcement (the workflow guidance the model reads) appears in sessions created after the host plugin mounted — i.e. after the first dsh restart following the install.
- The six workflow skills load with the preset and are invocable via the
skilltool in those sessions.
Development
pnpm install
pnpm test # preset-sync unit tests
pnpm run validate # structural validation of the bundled preset
pnpm run sync # sync the preset into ~/.dsh/.agent-presets
pnpm run verify # byte-compare the bundled preset against the install
License
MIT. The preset composition is adapted from the DeepSeek Harness built-in cordis preset (MIT); the workflow methodology is adapted from obra/superpowers (MIT); the preset-sync host-plugin pattern follows @linxin666/dsh-liangshen (Apache-2.0). See presets/engineering-workflow/NOTICE.
Links
More in this category
loopx-project/loopx#dsh-loopx-plugin★ 6694
LoopX, a provider-neutral, local-first state kernel and control plane for long-horizon agents: keeps Goal, Todo, gate, evidence, quota, recovery, and handoff state above DeepSeek Harness, while the plugin bootstraps the CLI and skills, admits bounded same-session continuation, and adds a loopback GoalBar for the exact bound loop.
Q00/ouroboros#integrations/dsh-plugin★ 6373
Config-only bundle that mounts Ouroboros through the DSH MCP client, exposing 36 interview, Seed, execution, evaluation, and evolution workflow tools in DSH.
chuspeeism/dashi-taskboard#deepseek-harness★ 3495
Embeds the active installed Codex Taskboard runtime in the DeepSeek Harness sidebar, using its launcher runtime descriptor instead of a fixed port.
NanmiCoder/dsh-agent-teams★ 2978
AgentTeams multi-agent teams.
EthanYoQ/AI-Novel-Writer#dsh-ai-novel-writer★ 2670
Installs a dedicated AI novel-writing preset and workbench: revisioned local project assets, a compact side drawer, and native approval-gated single-file changes.
shanliuling/dsh-image-gen★ 1576
Native conversational image generation for DeepSeek Harness: ask the agent to create an image, and it handles generation and keeps the result directly in the conversation.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.