界面已本地化;技能介绍与说明文档按所选语言由 AI 自动翻译,代码、命令、路径和专业术语保留原文,以原作者文档为准。
探索技能
开发与工程 · SKILL.md

generate

generate

【开发与工程】Regenerate .codex/agents/*.toml from agenteam.yaml and plugin defaults.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准

技能说明

GitHub

name: generate description: Regenerate .codex/agents/*.toml from agenteam.yaml and plugin defaults.

AgenTeam Generate

Regenerate Codex-native agent files from the current configuration.

When to Use

  • After editing agenteam.yaml manually
  • After adding or removing roles
  • After updating the plugin (new default role templates)
  • To verify generated agents match the config

Process

1. Validate Config

python3 <runtime>/agenteam_rt.py roles list

If this fails, the config has errors — show them and stop.

Role models should normally inherit the platform default. For any explicit pin, compare it with codex debug models; warn about unavailable or deprecated entries without rejecting values solely because a static AgenTeam list is old.

2. Generate Agents

python3 <runtime>/agenteam_rt.py generate

3. Report Results

Show what was generated:

Generated agents:
  .codex/agents/architect.toml      (updated)
  .codex/agents/dev.toml    (updated)
  .codex/agents/reviewer.toml       (updated)
  .codex/agents/qa.toml    (updated)

If custom roles were included, highlight them:

  .codex/agents/security_auditor.toml  (custom role)

4. Verify

Optionally show a summary of each generated agent's key fields (name, model, can_write, participates_in) for user verification. Display an omitted model as platform default, not as a missing setting.

If the user regenerated agents because roles are missing from the @ picker, tell them to confirm the files now exist under .codex/agents/ in the current project, then open a new thread or restart Codex so it reloads workspace agents.

我的收藏