需要先装好目标编码 agent 本体，Autoprompt 只是在其上安装 skill；Python 侧要单独装 PyYAML，否则环境检查不通过；原生 Windows 上 Codex 需要管理员权限运行控制器并预先初始化官方 Codex Windows 沙箱；它不会后台自动启动，必须显式执行 autoprompt activate 才会运行
autoprompt-skill
SpielewoyMIT活跃维护给编码智能体加装的自检技能，在自主编码任务中自动验证和纠错，把失败率降低约 45%。
Autoprompt is a coding-agent skill that cuts failures by 45% on agentic coding tasks.
v2.1.0发布于 2026-10-04 09:14 UTC1,214star68fork同步于 2026-11-07 07:32 UTC
npm install -g https://github.com/Spielewoy/autoprompt-skill/releases/download/v2.1.0/autoprompt-skill-2.1.0.tgz
autoprompt 硬件要求按 README 说明
README 未说明显存或 GPU 要求；运行依赖 Node.js 20+、Python 3.11+（含 PyYAML）和 Bash 4.3+，原生 Windows 走 AppContainer 隔离，无需 WSL 或虚拟机
给编码 agent 加一层自审工作流，通过复查、修复、再验证把任务失败率降低约 45%
给编码智能体加装的自检技能，在自主编码任务中自动验证和纠错，把失败率降低约 45%。
核心能力
- 按规划、执行、独立检查分层，避免同一 agent 自审自批
- 提供 direct、light、roadmap 三种工作路径，可按任务复杂度选择
- 用 --concurrency 控制并行子 agent 数量，可选 tokensaver 或 wide
- 支持 11 种编码 agent，包括 Claude Code、Codex、OpenCode 等
- 用 autoprompt doctor --strict 检查各 provider 安装是否可用
容易踩的坑
内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README
- v2.1.0v2.1最新正式版2026-10-04 · 2026-10-04 09:14 UTC
Native Windows runtime support without WSL2 or a VM.
- Fix OpenCode npm executable shims and PowerShell argument forwarding (#28).
- Add native Windows admission, including Claude Code 2.1.270 (#27).
- Complete platform fixes and x64/ARM64 verification across Windows, macOS, and Linux.
- Isolate Windows Codex credentials and enforce activation-bound network and process cleanup.
Windows Codex requires an administrator-started controller and an initialized official sandbox. Codex checks cover activation, sandbox enforcement, and cleanup; live-model missions were not tested.
Thanks to @Oscar-Williams for PR #29. His original authored commit is preserved in
main.Install:
npm install -g https://github.com/Spielewoy/autoprompt-skill/releases/download/v2.1.0/autoprompt-skill-2.1.0.tgz - v2.0.0v2.02026-09-09 · 2026-09-09 21:25 UTC
查看更新说明
Autoprompt v2 brings the Codex v2 workflow to all 11 supported providers, including Grok Build and Hermes, with model and effort controls, bounded subagents, and installer lifecycle support.
Download the package or your platform’s installer kit below. The standalone installers fetch this exact v2 release. Installation instructions and diagrams are updated in English, Chinese, Korean, Spanish, and Arabic.
Fresh install/uninstall checks pass for all 11 providers on Linux, and the final Codex native credential-isolation checks pass. Earlier eleven-provider live acceptance is retained, but the final Codex cheap-model runs passed the task test without achieving full checker acceptance: contradictory model-written probes or an invalid final checker envelope produced DONE_WITH_VERIFICATION_LIMITATIONS. Full live acceptance of the final Codex artifact is not established. The Windows and macOS downloads are installer kits; native runtime verification remains Linux-only.
Thanks to @johnatag, @rollingdice, @AincradBot, @lunar-me, @fatinghenji, @c8dhjp4tyv-bit, and @Alexis-Fiolleau-LaPoste-BGPN for the fixes, reports, and provider proposals.
- v1.0.4Autoprompt Skill 1.0.42026-08-21 · 2026-08-21 16:29 UTC
查看更新说明
Autoprompt Skill 1.0.4
Autoprompt turns one explicit goal into a closed plan, build, test, review, repair, and verification loop.
Highlights
- 45% fewer failures in the published Terminal-Bench 2.1 run: 29 failures fell to 16
- State-aware CLI scans nine coding agents, reports installed versions, and offers install, update, repair, doctor, and uninstall flows
- Older Codex installs are detected and updated in place
- Explicit invocation keeps the orchestration loop isolated from ordinary coding requests
- Native provider packages for Claude Code, Codex, OpenCode, Kilo Code, VS Code, Prime Agent, Oh My Pi, DeepSeek Harness, and Reasonix
- Deterministic English, Chinese, Korean, Spanish, and Arabic documentation visuals
Install
npm install -g autoprompt-skill autopromptSupported coding agents
- Claude Code
- Codex
- OpenCode
- Kilo Code
- VS Code
- Prime Agent
- Oh My Pi
- DeepSeek Harness
- Reasonix
Release assets
- npm archive for package managers and offline installation
- Windows kit with PowerShell bootstrap
- Linux kit with Bash bootstrap
- macOS kit with Bash bootstrap
- standalone bootstrap scripts
- SHA-256 checksums for every downloadable asset
Requirements
Node.js 20 or newer, Python 3.11 or newer with PyYAML, and Bash 4.3 or newer on Linux or macOS.
See the README for provider versions, custom roots, model routing, and benchmark evidence.
- v1.0.3Autoprompt Skill 1.0.32026-08-20 · 2026-08-20 12:48 UTC
查看更新说明
Autoprompt Skill 1.0.3
Autoprompt turns one explicit goal into a closed plan, build, test, review, repair, and verification loop.
Highlights
- 45% fewer failures in the published Terminal-Bench 2.1 run: 29 failures fell to 16
- State-aware CLI scans six coding agents, reports installed versions, and offers install, update, repair, doctor, and uninstall flows
- Older Codex installs are detected and updated in place
- Explicit invocation keeps the orchestration loop isolated from ordinary coding requests
- Native provider packages for Claude Code, Codex, OpenCode, Kilo Code, VS Code, and Prime Agent
- Deterministic English, Chinese, Korean, Spanish, and Arabic documentation visuals
Install
npm install -g autoprompt-skill autopromptSupported coding agents
- Claude Code
- Codex
- OpenCode
- Kilo Code
- VS Code
- Prime Agent
Release assets
- npm archive for package managers and offline installation
- Windows kit with PowerShell bootstrap
- Linux kit with Bash bootstrap
- macOS kit with Bash bootstrap
- standalone bootstrap scripts
- SHA-256 checksums for every downloadable asset
Requirements
Node.js 20 or newer, Python 3.11 or newer with PyYAML, and Bash 4.3 or newer on Linux or macOS.
See the README for provider versions, custom roots, model routing, and benchmark evidence.
- v1.0.2Autoprompt Skill 1.0.22026-08-19 · 2026-08-19 22:41 UTC
查看更新说明
Autoprompt Skill 1.0.2
Autoprompt turns one explicit goal into a closed plan, build, test, review, repair, and verification loop.
Highlights
- 45% fewer failures in the published Terminal-Bench 2.1 run: 29 failures fell to 16
- State-aware CLI scans six coding agents, reports installed versions, and offers install, update, repair, doctor, and uninstall flows
- Older Codex installs are detected and updated in place
- Explicit invocation keeps the orchestration loop isolated from ordinary coding requests
- Native provider packages for Claude Code, Codex, OpenCode, Kilo Code, VS Code, and Prime Agent
- Deterministic English, Chinese, Korean, Spanish, and Arabic documentation visuals
Install
npm install -g autoprompt-skill autopromptSupported coding agents
- Claude Code
- Codex
- OpenCode
- Kilo Code
- VS Code
- Prime Agent
Release assets
- npm archive for package managers and offline installation
- Windows kit with PowerShell bootstrap
- Linux kit with Bash bootstrap
- macOS kit with Bash bootstrap
- standalone bootstrap scripts
- SHA-256 checksums for every downloadable asset
Requirements
Node.js 20 or newer, Python 3.11 or newer with PyYAML, and Bash 4.3 or newer on Linux or macOS.
See the README for provider versions, custom roots, model routing, and benchmark evidence.
用户点评
登录用户的真实使用记录，可按版本和硬件筛选共 0 条
还没有人点评。第一个说出真实体验的人最有价值。
Contents
Install · Benchmarks · Invocation · Run controls · Workflow · Agents · Examples · FAQ · License
Install
Use the CLI below, or download an installer from GitHub Releases.
1. Install the CLI
npm install -g https://github.com/Spielewoy/autoprompt-skill/releases/download/v2.1.0/autoprompt-skill-2.1.0.tgz
2. Launch the installer
autoprompt
3. Install
Choose your coding agent, confirm its path, and install. N means enter another path.
For another CLI or IDE, choose Custom coding agent and use the compatibility guide.
git clone https://github.com/Spielewoy/autoprompt-skill
cd autoprompt-skill
npm install -g .
autoprompt
Requirements
- Node.js 20+
- Python 3.11+ available as
python3orpython, with PyYAML - Bash 4.3+ on macOS or Linux
- Windows native provider commands use the installed provider runtime plus the platform sandbox and owned-process controls; WSL and a hypervisor are not required for native paths
- Native Windows Codex additionally requires an administrator-run Autoprompt controller and an initialized official Codex Windows sandbox. Setting Codex
windows.sandbox="elevated"selects that sandbox mode; it does not elevate the Autoprompt controller. - Git only for the GitHub checkout method
Support
| Status | Coding agent | Tested version | Key |
|---|---|---|---|
| Working | Claude Code | 2.1.263 | claude |
| Working | Codex | 0.148.0 | codex |
| Working | OpenCode | 1.18.29 | opencode |
| Working | Kilo Code | 7.5.15 | kilo |
| Working | VS Code | 1.136.1 | vscode |
| Working | Prime Agent | 0.7.2 | prime |
| Working | Oh My Pi | 18.1.14 | omp |
| Working | DeepSeek Harness | 0.1.2-rc.1 | deepseek |
| Working | Reasonix | 1.30.0 | reasonix |
| Working | Hermes Agent | 0.21.1 | hermes |
| Working | Grok Build | 1.0.13 | grok |
These versions passed Linux runs. Model and platform availability varies by provider.
Native Windows execution uses AppContainer isolation and Windows Job process ownership, with Node 20 or 24 on x64 or ARM64. Native macOS execution uses the system sandbox and owned-process controls. Windows command tools include the bundled Bash/MSYS/Node runtime; native paths do not require WSL or a VM.
The ten non-Codex providers can run their local capability checks on Linux,
Windows, and macOS. All eleven checks must pass for the installed runtime before
a mission starts; missing, failed, or skipped checks block activation. Codex uses
its separate canonical validation route. On Windows, that route creates a fresh
offline account and an own-SID Windows Filtering Platform policy for each
activation. Their activation-bound lease is retained for resume. Permanent
cleanup removes the account and policy only after owned Jobs drain; uncertain
verification or cleanup fails closed and retains recovery records. The Linux
version table above does not
certify every provider/platform combination. Installer success and
doctor --strict are prerequisite checks; local-canary-required means native
validation is still pending. See configured Linux runtimes
when a native path is unavailable.
Check, update, or remove an installation
- Check every detected installation:
autoprompt doctor --strict - Check one provider:
autoprompt doctor PROVIDER --strict - Update or repair:
autoprompt, then choose an installed provider - Uninstall interactively:
autoprompt uninstall - Uninstall one provider:
autoprompt uninstall PROVIDER - Show every command:
autoprompt help
Replace PROVIDER with a key from the support table, such as claude, codex, or prime.
Benchmarks
These are version 1 benchmarks. Version 2 benchmarks will follow.
| Track | Solved | Score | Failed |
|---|---|---|---|
| OpenCode | 60/89 | 67.42% | 29 |
| OpenCode + Autoprompt | 73/89 | 82.02% | 16 |
| Change | +13 solves | +14.61 points | 45% fewer |
DeepSeek's 82.7% used its own test setup, so it is a reference point, not a comparable third run. Read the setup and evidence boundaries, or request another benchmark.
Timing and token logs were not retained, so these are planning estimates based on user experience reports, not measured benchmark results. The measured result was 29 to 16 failures (45% fewer) in this run, which translates to about 2x fewer mistakes. Note: for very small tasks, this may differ significantly.
Anatomy of an invocation
autoprompt activate PROVIDER --target /absolute/project -- "<goal>"
| Part | What it does |
|---|---|
PROVIDER | A key from the support table, such as claude, codex, or grok. |
--target | The project to work on. Omit it to use the current directory. |
-- | Separates launcher options from the request. |
<goal> | The result you want, constraints, and how to check success. |
path= | Optional auto, direct, light, or roadmap, before the quoted goal. See work paths. |
autoprompt activate codex -- path=light "add retries and test the edge cases"
Run controls
The same controls apply to all eleven providers. Custom model setup
| Control | What it does |
|---|---|
--concurrency tokensaver | Runs at most six subagents at once. |
--concurrency wide | Starts ready, independent work up to the host limit. |
--concurrency custom --max-subs N | Sets your own concurrency limit. |
configure PROVIDER --agents off | Uses the provider's configured model. |
configure PROVIDER --agents MODEL | Selects one model. Add --effort LEVEL where supported. |
configure PROVIDER --agents auto --model-map FILE | Chooses from a measured model registry. A comma-separated model list also requires --model-map. |
Pass concurrency controls after --, before the quoted goal:
autoprompt activate codex -- --concurrency custom --max-subs 4 "add retries and tests"
autoprompt configure claude --agents provider/model --effort low
How it works
The agents
Examples
| Goal | Prompt |
|---|---|
| Fix | autoprompt activate claude -- "fix the registration race and add a regression test" |
| Build | autoprompt activate codex -- --concurrency wide "build the booking flow from API to checkout" |
| Research | autoprompt activate hermes -- "compare job queues against this codebase and recommend one" |
| Limit parallel work | autoprompt activate grok -- --concurrency custom --max-subs 4 "migrate every model" |
Run these commands from your project, or supply --target /absolute/project before --.
FAQ
No. Give it a clear goal, constraints, and success criteria. Autoprompt handles the execution loop, so you do not have to prompt every step. Details
It can scope, implement, test, review, repair, and verify a goal. It stops for choices that change the result, actions that need your authority, or blockers it cannot safely resolve. Details
The layers separate coordination, management, execution, and independent judgment. That separation keeps one agent from planning, approving, and verifying its own work. Details
path=auto selects a route for the task. direct starts focused work, light adds a short plan, and roadmap organizes dependent work before execution. Every path includes independent verification. Details
--concurrency and --max-subs set parallel work limits. configure --agents selects models, and path= selects how work is planned and coordinated. Details
Because it changes cost, time, and workflow. Start it explicitly with autoprompt activate PROVIDER -- "<goal>".
License
MIT. Copyright 2026 Spielewoy.
Community: Contributors, Contributing, Code of Conduct, Security, and Support.
Contributors
johnatag · rollingdice · AincradBot · lunar-me · fatinghenji · c8dhjp4tyv-bit · Alexis-Fiolleau-LaPoste-BGPN · Oscar-Williams