Generate a final plan by aggregating parallel proposals from multiple AI model labs.
npx skills add https://github.com/drivelineresearch/moa-x --skill mixture-of-agents
Please help me install this Agent Skill. Skill: mixture-of-agents Source: https://github.com/drivelineresearch/moa-x/tree/main/harness Command: npx skills add https://github.com/drivelineresearch/moa-x --skill mixture-of-agents
This Skill facilitates collaborative planning by leveraging an ensemble of models from different labs, enabling comprehensive analysis and refined outcomes.
/mixture-of-agents --spec ./docs/cache-layer-spec.md
High-intent search queries and answers about installing and using this skill.
Cross-lab ensemble planning uses models from OpenAI, Anthropic, Zhipu, and Moonshot to generate parallel proposals, perform broadcast refinement, and aggregate them into a final repository-based plan.
Broadcast refinement is a process where initial planning proposals are independently reviewed by different models to verify logic, resolve conflicts, and ensure robust architecture decisions.
Yes, repository-based ensemble planning requires external CLI tools like codex, opencode, claude, and cursor, along with Python execution to manage the parallel proposal workflow.
Yes, this approach is tailored for repo-grounded planning, making it ideal for complex software architecture decisions where multiple model perspectives are crucial for evaluation.
You can initiate planning by passing a specification file path as an argument, allowing the ensemble to read the requirements and generate aggregated proposals based on your repository.
korchasaExecute code prompts across Claude Code, OpenCode, Cursor, and Codex and compare outputs.
dlantsConvert vague requirements into structured implementation plans saved to plans/<planName>.md.
khapn-2933Analyze repository contents to generate technical and business upsale proposals.
quangdang46Plan multi-model software project architectures before coding.
