b3ehive
от weiyangzen
Coding Agents & IDEsКратко
Enable competitive code generation where three isolated AI agents implement the same functionality, evaluate.
Описание
Что делает этот навык
Three AI agents independently implement the same coding task with different optimization priorities (simplicity, speed, robustness), then cross-evaluate each other using weighted numeric scores across five dimensions. The highest-scoring or hybrid solution is delivered alongside a comparison report and decision rationale.
Rather than accepting a single AI-generated solution, you get three independently developed implementations with objective cross-evaluation scores, making trade-offs explicit and selection data-driven.
Когда его использовать
- Choosing between a readable vs performant sorting implementation
- Getting three independent solutions to a concurrency problem before picking one
- Generating a documented comparison for a code review or architecture decision
- Stress-testing edge case coverage across independent implementations
- Deciding whether to optimize for speed or stability on a new API endpoint
Пример сценария
User asks: Implement a thread-safe rate limiter
- 1
Spawn three agents in parallel: Agent A focuses on simplicity, Agent B on speed, Agent C on robustness
- 2
Each agent writes a complete runnable implementation with tests and a SUMMARY.md
- 3
Each agent evaluates the other two implementations across simplicity, speed, stability, corner cases, and maintainability
- 4
Each agent produces a SCORECARD.md with numeric scores for all three solutions
- 5
Select the winner based on score margins and output the final solution, COMPARISON_REPORT.md, and DECISION_RATIONALE.md
A runnable winning implementation with a full comparison report and transparent decision rationale
Установка
clawhub install weiyangzen/b3ehiveВыполните команду CLI или вставьте промпт в Claude Code, Cursor, Codex — любой кодинг-агент, которым пользуетесь.