Autoprompt is an MIT-licensed coding-agent skill for developers and teams working in agentic coding environments who want fewer failed runs, adding a review, fix, and recheck workflow to supported coding agents such as Claude Code, Codex, and OpenCode.

What it is

Autoprompt is distributed as autoprompt-skill, a JavaScript package on npm and installable through the CLI. It belongs to Developer Tools / Version Control & Collaboration and carries topics such as agent-orchestration, agent-skills, agentic-workflows, ai-agents, ai-coding, automated-testing, and code-review. It installs into coding agents rather than standing alone as an agent or IDE. The README lists working support for eleven coding agents, including Claude Code, Codex, OpenCode, Kilo Code, VS Code, Prime Agent, Oh My Pi, DeepSeek Harness, Reasonix, Hermes Agent, and Grok Build, with tested versions that passed Linux runs; model and platform availability varies by provider.

The problem is failed agentic coding tasks. Autoprompt cuts failures by 45% by reviewing, fixing, and rechecking its work. In the measured OpenCode comparison, OpenCode solved 60/89 tasks (67.42%) and failed 29; OpenCode plus Autoprompt solved 73/89 (82.02%) and failed 16, a change of +13 solves, +14.61 points, and 45% fewer failures. The measured result was 29 to 16 failures, about 2x fewer mistakes. These are version 1 benchmarks; version 2 benchmarks will follow. DeepSeek's 82.7% used its own test setup, so it is a reference point, not a comparable third run. Autoprompt does not replace the coding agent itself; it replaces unreviewed agent output with reviewed, fixed, and rechecked output.

Key capabilities

• Installs with npm install -g https://github.com/Spielewoy/autoprompt-skill/releases/download/v2.0.0/autoprompt-skill-2.0.0.tgz, then runs autoprompt to choose a coding agent, confirm its path, and install.
• Supports provider keys claude, codex, opencode, kilo, vscode, prime, omp, deepseek, reasonix, hermes, and grok.
• Runs a review, fix, and recheck workflow, measured at 45% fewer failures in the OpenCode comparison.
• Checks installations with autoprompt doctor --strict or autoprompt doctor PROVIDER --strict.
• Removes installations with autoprompt uninstall or autoprompt uninstall PROVIDER, and shows every command with autoprompt help.
• Invokes with autoprompt activate PROVIDER --target /absolute/project -- " ".
• Supports custom CLIs or IDEs via Custom coding agent and docs/guides/custom-agent-compatibility.md.

Who uses it and how

• Developers using supported agents such as Claude Code, Codex, OpenCode, Kilo Code, VS Code, Prime Agent, Oh My Pi, DeepSeek Harness, Reasonix, Hermes Agent, or Grok Build who want review, fix, and recheck in the agent workflow.
• Teams standardizing on a provider key such as claude, codex, or prime that need check, update, repair, and uninstall commands for each installation.
• Self-hosters on macOS or Linux who provide Node.js 20+, Python 3.11+ with PyYAML, and Bash 4.3+, with Git needed only for the GitHub checkout method.
• Users evaluating agentic coding performance against the Terminal-Bench 2.1 setup in docs/benchmarks/terminal-bench-2.1.md or requesting another benchmark through the issue tracker.
• Users integrating a custom CLI or IDE by choosing Custom coding agent and following the compatibility guide.

Getting started

Install the CLI with npm install -g https://github.com/Spielewoy/autoprompt-skill/releases/download/v2.0.0/autoprompt-skill-2.0.0.tgz, then run autoprompt to choose an agent, confirm its path, and install. From source, run git clone https://github.com/Spielewoy/autoprompt-skill, cd autoprompt-skill, npm install -g ., and autoprompt.

How it compares

Autoprompt does not replace Claude Code, Codex, OpenCode, Kilo Code, VS Code, Prime Agent, Oh My Pi, DeepSeek Harness, Reasonix, Hermes Agent, or Grok Build; it installs as a skill on top of them. It sits alongside those agentic coding tools in Developer Tools / Version Control & Collaboration, adding review, fix, and recheck while the underlying agent remains the execution environment.

When to use it — and when not to

Use Autoprompt when a supported coding agent is in place and the expected trade-off of about 3x the time and 2x the tokens is acceptable, with version 1 benchmark results as the available evidence. Do not pick it if version 2 benchmarks are required before adoption, if that time and token cost is too high, or if the coding agent or IDE is unsupported and cannot follow the custom-agent compatibility guide. The README states that timing and token logs were not retained, so those figures are planning estimates based on user experience reports rather than measured benchmark results, that very small tasks may differ significantly, and that a self-hoster must provide Node.js 20+, Python 3.11+ with PyYAML, and Bash 4.3+ on macOS or Linux.

npm install -g https://github.com/Spielewoy/autoprompt-skill/releases/download/v2.0.0/autoprompt-skill-2.0.0.tgz

autoprompt

git clone https://github.com/Spielewoy/autoprompt-skill
cd autoprompt-skill
npm install -g .
autoprompt

autoprompt activate PROVIDER --target /absolute/project -- "<goal>"

autoprompt activate codex -- path=light "add retries and test the edge cases"

autoprompt activate codex -- --concurrency custom --max-subs 4 "add retries and tests"
autoprompt configure claude --agents provider/model --effort low