Our review
This skill asks a coordinator to identify genuinely separate problem domains, give each agent a bounded prompt, and integrate the results after they finish. It uses examples from failing test groups and stresses clear expected outputs. The goal is to avoid several agents editing the same shared state or pursuing overlapping theories.
Use it when multiple failures can be investigated without shared files or ordered dependencies. Skip it when a single root cause is likely, when tasks mutate the same files, or when coordination costs exceed the work. Parallel results still need a joint review and fresh verification. The instruction assumes available subagent tools and sufficient budget, so adapt it to the actual environment.
This folder is only SKILL.md; it contains no bundled script or direct network call. Dispatching agents can cause whatever local or external actions their prompts authorize, so define permissions explicitly for each task. Read the source before adopting the example commands or prompt structure.
For installation, open the linked maintainer folder and read its current SKILL.md, license, and supporting files. Copy the complete folder to a skills location your agent documents, or use an installer your agent supports. Begin with a small, low-risk task and confirm that the agent loaded the intended version. Our install guide covers those checks.
Source and license
Maintainer: Jesse Vincent / obra. Checked: October 18, 2026. The maintainer controls this repository; the current files may differ from the version we reviewed.