victorzhong0110/dsh-outcome-loop

DeepSeek Harness (DSH) 任务结果账本与验收插件:机械证据验收、默认零额外token、数据本地持有。本地优先的任务结果账本与验收插件:基于证据的验证、零额外模型成本、用户拥有的附属数据。

Project Overview项目介绍

dsh-outcome-loop is a task outcome ledger and acceptance plugin built exclusively for DeepSeek Harness. To install it, users first clone the repository from GitHub, run pnpm install and pnpm build to generate a compressed plugin tarball, then install it into a DSH profile via the DSH CLI plugin add command. The plugin requires the storageDomain service, which is not included in the official dsh-base bundle, so users on bare headless profiles need to manually add the dependency if they do not have it already and update their cordis patch configuration before the plugin will boot correctly.

Once installed, the plugin runs as a passive session observer that tracks task progress without adding extra costs or changes to the LLM workflow. Users interact with it via the /outcome command set, which lets them create new task contracts, add multiple types of acceptance criteria, run automated verification, check current status, and record their own disposition of the task. It is designed for any DSH user that wants a clear, auditable record of what was done, whether they are a developer building features or a regular user working through multi-step tasks.

All outcome data is stored locally on the user’s machine in a separate sidecar storage domain, never in session logs or telemetry, and export is an explicit two-step process that requires user approval after preview. The plugin does not make any network calls or LLM calls by default, and it only stores structured facts about the task, never full prompts, source code or message bodies. It is released under the permissive MIT open source license, and it is developed and verified against DSH 0.1.0-rc.7, with a compatibility matrix available to check for breaking changes as DSH evolves.

这是一个专门为DeepSeek Harness(DSH)开发的原生插件,功能是任务结果记账与验收,核心目标是区分“模型说完成”和“任务实际完成”,基于可机械验证的证据判断任务状态。它会把一次DSH会话的目标、约束、验收标准、执行证据、用户反馈、token消耗成本和最终结果整理成可复核的任务记录,所有数据默认存储在用户本地的DSH存储后端,用户完全掌控自己的所有数据。

用户可以通过/outcome系列命令创建任务合约、添加手动规则、命令输出、测试报告、产物文件等不同类型的验收标准,运行验证后查看任务状态,再自行决定接受、拒绝、修改还是放弃任务。它默认不产生额外的LLM调用成本,也不会修改系统提示或给模型增加额外可见工具,适合需要留存可审计任务记录、明确任务交付结果的DSH开发者和普通用户使用。

这个插件依赖DSH的storageDomain服务,官方基础dsh-base包不提供该服务,需要用户手动添加对应依赖并修改配置,否则会启动失败。项目采用MIT许可开源,默认零网络请求、零模型调用,数据单独存储在侧边栏域,不会混入会话日志或遥测数据,导出需要用户手动两步确认,隐私默认配置偏向严格保护。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 1 warning1 项注意
  • Only 2 stars - very few users, little community feedback星标只有 2,几乎没人在用,遇到问题缺少社区反馈
DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

  • DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
  • External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时(Electron / Python / Docker 等)是否齐备
  • Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突:命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

  • Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致,是否归档或长期停更
  • Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
  • curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
  • Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
  • Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需(filesystem / network / shell / clipboard)
  • Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限,以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照,不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装(适合开发者)

CLI Install命令行安装

dsh plugin --profile web add github:victorzhong0110/dsh-outcome-loop

把 victorzhong0110/dsh-outcome-loop 加入你的 DSH 配置(web profile)即可启用。

READMEREADME

dsh-outcome-loop

Task outcome ledger & acceptance plugin for DeepSeek Harness (DSH) · 任务结果账本与验收插件

Know whether a task was actually completed — with near-zero extra tokens and mechanically re-checkable evidence — and keep the result as a portable, user-owned record.

dsh-outcome-loop is a local-first, user-owned, vendor-neutral task outcome ledger and acceptance plugin for DSH. It organizes one DSH session's goal, constraints, acceptance criteria, execution evidence, user feedback, cost, and final result into a re-auditable task record.

  • Stops "the model said done" from being mistaken for "the task is done": acceptance is based on mechanical evidence — tests, build, lint, exit codes, file state, diagnostics, git scope;
  • Zero extra model cost by default: no extra LLM calls, no model-visible tools, no system-prompt injection;
  • Local by default, offline by default: all result data lives in your own DSH storage backend;
  • User owns and controls the data: inspect, delete, or export — with a preview + redaction gate before any export;
  • Separates success, failure, unknown, stale evidence, user acceptance and user abandonment — an unknown is never silently converted into a success.

Interview oral (3 minutes)

One-page whiteboard script (no ADR binder): docs/ORAL-3MIN.md. Covers problem → data/control plane → non-negotiables → proof we are not cosplaying Datadog.

Quick start

Install

# From source
pnpm install
pnpm build
pnpm pack   # produces dsh-outcome-loop-<version>.tgz

# Install into a DSH profile
dsh plugin --profile <name> add ./dsh-outcome-loop-0.1.0-beta.8.tgz

Prerequisite — storageDomain (real-host verified): the plugin requires the storageDomain service, which the official dsh-base bundle does not provide (only upper bundles such as @deepseek-ai/dsh-web-app do; the web profile already has it). On a bare/headless profile, add it once:

dsh plugin --profile <name> add @deepseek-ai/dsh-storage-domain@0.1.0-rc.8
# then append to ~/.dsh/profiles/<name>/cordis.patch.yml:
# - insert:
#     - id: storage
#       name: "@deepseek-ai/dsh-storage"
#     - id: storage-json
#       name: "@deepseek-ai/dsh-storage-json"
#       config:
#         root: !!js dshHomePath("storages")
#     - id: storage-domain
#       name: "@deepseek-ai/dsh-storage-domain"
#       config:
#         backend: json

Without it the profile boot fails loud with waiting for service: storageDomain (fail-loud by design; see COMPATIBILITY.md §4 for the verified lifecycle: add → dump-config → boot → real /outcome commands → restart-read → uninstall).

Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要,完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →

← 上一个 Prev dsh-safemode-profile 下一个 Next pocket-dsh →