Menu
Design controlled coding-agent tasks for measuring developer-tool agent experience.
The review-first prompt is selected by default. You can switch to a direct command or download a local copy.
Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
npx skills add https://github.com/shenli/devtool-ax-kit --skill ax-task-designThe command stays on one line. Scroll horizontally to inspect it before copying.
Prefer a local copy? Download the files currently available to SkillsMP.
Showing SKILL.md
| name | ax-task-design |
| description | Design controlled coding-agent tasks for measuring developer-tool agent experience. |
Create a small, realistic task with a frozen starting fixture, explicit allowed side effects, an independent verifier, and a clear stop condition. Separate the task prompt from the verifier. Define what the agent may use (raw docs, CLI, Skill, MCP, or plugin) so surface comparisons are attributable. Prefer disposable local or test resources; never require production credentials.
Make the task legible from the agent's available context: state where the source of truth, logs, acceptance criteria, and recovery instructions live. Include at least one boundary or restart assertion when the task has durable state.
Record the fixture hash, prompt, harness/model, repetitions, and human interventions. Keep the task narrow enough that a failure can be reproduced and reported to the tool owner.
