‹ 首页

dev-learner

@evolution-foundation · 收录于 5 天前 · 上游提交 2 个月前

Meta-skill — extract reusable skills from the current conversation. Identifies patterns that appeared 3+ times and proposes them as new skills. Pairs with skill-creator builtin.

适合你,如果经常在对话中重复做类似操作,想将其固化为技能

/ 通过 npx 安装 校验哈希
npx oh-my-skill add evolution-foundation/evo-nexus/dev-learner
/ 通过 bash 安装
curl -fsSL https://oh-my-skill.com/install.sh | bash -s -- evolution-foundation/evo-nexus/dev-learner
/ 已经装过?验证本机副本,不用重装
npx oh-my-skill verify evolution-foundation/evo-nexus/dev-learner
安装目标可用 --agent / --scope 或 --to 明确指定;省略时只会在唯一已存在的 agent 目录上自动选择,零命中或多命中会停止并提示。content_hash 缺失或不一致均拒装。
509GitHub stars
~541上下文体积 · 单文件
索引托管

怎么用

商店整理自技能原文 · 版本 7f5dd76 · 表述以原文为准
它做什么

装上后,Claude会扫描当前对话,找出重复3次以上的工作流、提示、输出或交接模式,并提议将模式提取为新技能,生成学习报告。

什么时候触发

当用户进行长对话,有重复3次以上的相同流程,或希望将手动重复的模式正式化为技能时触发。常用于回顾EPIC后或技能复盘。

装好后可以这样说
Claude会扫描并列出检测到的模式。
Claude会识别重复操作并提出新技能建议。
技能原文 SKILL.md作者撰写 · Apache-2.0 · 7f5dd76

Dev Learner

Derived from oh-my-claudecode (MIT, Yeachan Heo). Adapted for the EvoNexus Engineering Layer.

Meta-skill: scan the current conversation for patterns that appeared 3+ times and propose extracting them as reusable skills. The "self-improving workspace" feedback loop.

Use When
  • A long session involved repeated workflows that aren't yet skills
  • You want to formalize a pattern you keep manually invoking
  • Post-EPIC retrospective: what should have been a skill but wasn't?
Do Not Use When
  • Single-pattern conversation → no extraction value
  • The pattern is too specific to one project → keep as agent memory
  • The pattern is already a skill → just document it better
Workflow
Phase 1 — Pattern Detection

Scan the conversation for:

  • Repeated workflows — same sequence of agents called 3+ times
  • Repeated prompts — similar instructions issued repeatedly
  • Repeated outputs — same artifact format produced multiple times
  • Repeated handoffs — same agent pair invoked together repeatedly
Phase 2 — Extraction

For each detected pattern:

  1. Name it (e.g., "PR review pipeline", "Debug investigation flow")
  2. Describe the trigger
  3. Document the steps
  4. Identify the agents involved
  5. Identify the inputs and outputs
Phase 3 — Proposal

Propose the new skill:

  • Name: suggested skill name (dev-{action} if engineering, {prefix}-{action} otherwise)
  • Trigger: when to invoke
  • Workflow: the steps
  • Pairs with: which agents/skills
Phase 4 — Creation (optional)

If user accepts, hand off to skill-creator (builtin) to actually generate the skill file.

Output
## Learner Report — {session topic}

### Patterns Detected
1. **{Pattern name}**
   - Occurrences: 4 times
   - Steps: ...
   - Suggested skill: `dev-{action}`

2. **{Pattern name}**
   - ...

### Skill Proposals
[Detailed proposal for each detected pattern]

### Recommendation
[Which to extract first, which to skip]
Pairs With
  • skill-creator (builtin) — to generate the actual skill file
  • create-agent (builtin) — if the pattern warrants a new agent instead
  • All 19 engineering agents (which patterns might be extracted from)
Anti-patterns
  • Over-extraction (every 2-occurrence sequence becomes a skill)
  • Under-extraction (missing genuine repeated patterns)
  • Naming conflicts (proposing names that collide with existing skills)
按 Apache-2.0 许可原样转载,未经改动 · 在 GitHub 查看 →

评论

登录即可评论;带「已验证安装」的,是发布者名下有本店的安装或持有记录。