AGENT SKILL

evolve-select

暂无简介

License 许可
MIT
最近更新
2026-06-13
Stars
★ 3
Forks
1

01 前言 Front Matter

id
evolve-select
title
Select Top-K Training Configurations

03 SKILL.md 正文

Select Top-K Training Configurations

Rank evaluated LLM training configurations and select the top-K as parents for the next generation.

Process

  1. Read the consolidated scores for the current generation
  2. Rank configurations by overall fitness (descending)
  3. Select the top-K configurations as parents
  4. Compare this generation's best to the all-time best
  5. Update best-candidate.json if a new best is found
  6. Advance the generation counter in evolution-state.json
  7. Write selection results for the crossover phase

Outputs

  • selection.json:
    {
      "generation": 0,
      "parents": [
        {
          "candidateId": "candidate-003",
          "fitness": 0.85,
          "specification": {
            "architecture": { "attentionType": "GQA", "layers": 32, "..." : "..." },
            "hyperparameters": { "learningRate": 2e-4, "..." : "..." },
            "dataStrategy": { "mix": { "code": 0.3, "..." : "..." }, "..." : "..." }
          }
        }
      ],
      "populationStats": { "mean": 0.7, "max": 0.85, "min": 0.55 }
    }
    
  • best-candidate.json:
    {
      "candidateId": "candidate-003",
      "generation": 0,
      "fitness": 0.85,
      "approach": "GQA architecture with code-heavy curriculum",
      "specification": { "architecture": {...}, "hyperparameters": {...}, "dataStrategy": {...} }
    }
    
  • evolution-state.json — generation counter incremented, status updated

Quality Criteria

  • Selection is strictly by fitness ranking
  • Best configuration is tracked across all generations (not just current)
  • Generation counter is advanced exactly by 1
  • Parent specifications are preserved in full (needed by crossover)

04 同仓库更多 Skills

openplaybooks-dev
openplaybooks-dev/converge

hello-checker

Use this skill when asked to "verify", "check", or "validate" a file's existence and content. It provides a reliable workflow for confirming files meet their spec.

Skill与智能体开发
★ 3
openplaybooks-dev
openplaybooks-dev/converge

ask-skill

暂无简介

Skill与智能体开发
★ 3
openplaybooks-dev
openplaybooks-dev/converge

html-review-artifact

Framework-wide guidance for crafting polished human-review artifacts as markdown or HTML

前端开发Skill与智能体开发
★ 3
openplaybooks-dev
openplaybooks-dev/converge

converge-planning

Design a Converge playbook end-to-end. Use when the user says "plan a project", "design a playbook", "decompose into tasks", "scaffold a Converge workflow", "onboard to Converge", or asks how to write TASK.md / playbook.yml. Two entry points: plan a new playbook or update an existing one. Hand off to `/converge-control` after producing the playbook structure.

Skill与智能体开发
★ 3
openplaybooks-dev
openplaybooks-dev/converge

converge-control

Operate a Converge playbook end-to-end — start runs, monitor progress, narrow failures with `--select`, inspect runtime state, validate definitions, surgically reset subtrees, and stop stuck or orphaned runs. Use this skill whenever the user says "run my playbook", "execute this", "monitor the run", "converge failed", "the run is stuck", "retry failures", "babysit this playbook", "what's the status?", "show me the gantt", or asks how to use `converge run` / `--select` / `inspect` / `doctor` / `clean` / `stop`. Also use it for `HTTP 401` / `Invalid API key` failures on the first task — those are environment-not-playbook problems — and for any post-run cleanup work. Hand off to `converge-planning` if the user wants to design or restructure the playbook itself.

开发工程师Skill与智能体开发
★ 3
openplaybooks-dev
openplaybooks-dev/converge

repair-taskts

Fix task.ts definition issues — Seed syntax, missing .build(), empty checks, import errors, TypeScript errors

开发工程师
★ 3