skills
Use when designing or optimizing skill files or skill behavior; creates, reviews, and improves reusable agent skills with strong trigger descriptions, concise workflows, progressive disclosure, and realistic evaluations.
- License 许可
- MIT
- 所属仓库
- amoai-tech/mdeai
- 最近更新
- 2026-11-13
- Stars
- ★ 2
- Forks
- 1
01 前言 Front Matter
- name
- skills
- description
- Use when designing or optimizing skill files or skill behavior; creates, reviews, and improves reusable agent skills with strong trigger descriptions, concise workflows, progressive disclosure, and realistic evaluations.
02 资源文件 1
- 参考文档references/full-prompt.md2.6 KB
03 SKILL.md 正文
SKILLS — Create & Improve Agent Skills
Goal Create concise, discoverable skills that measurably improve agent performance.
Define the skill
- One skill = one reusable job.
- Define what it does, when it should trigger, when it should not, expected output, and mutation authority.
- Split skills when triggers, permissions, or outcomes differ.
Metadata
name: lowercase letters/numbers/hyphens; short and specific.description: state what it does + when to use it.- Put trigger guidance in the description, not deep in the body.
Keep it lean Target 40–80 lines, ~100 max by default. Keep SKILL.md to:
workflow → decisions → critical rules → resources → output
Move repeated deterministic work to scripts/, long guidance to references/, templates to assets/.
Match detail to risk
- High freedom: research/reviews.
- Medium freedom: normal implementation.
- Low freedom: migrations, destructive writes, release procedures. Use exact steps only when deviation is risky.
Build from evidence Before adding detail:
- create ~3 realistic evals
- establish baseline behavior
- write the minimum instructions needed
- rerun and compare
- remove instructions that add no value For new skills compare with no-skill baseline; for revisions compare with the previous version when practical.
Test triggering Use realistic:
- should-trigger prompts
- near-miss should-not-trigger prompts Check adjacent skills do not compete incorrectly.
Improve from failures Look for:
- missed triggers
- ignored rules
- wasted work
- repeated helper logic
- stale tools/examples Generalize fixes; do not overfit to one example.
For remaining steps, checks, and examples, see full prompt.
04 同仓库更多 Skills
mde-pr-review-supabase
Use when reviewing MDE Supabase SQL, Auth and RLS changes in a pull request.
mde-pr-review-stripe
Use when reviewing MDE Stripe payments and webhooks changes in a pull request.
mde-pr-review-nextjs
Use when reviewing MDE Next.js App Router changes in a pull request.
mde-pr-review-mastra
Use when reviewing MDE Mastra agent, tool and workflow changes in a pull request.
mde-pr-review-maps
Use when reviewing MDE Google Maps and Places changes in a pull request.
mde-pr-review-copilotkit
Use when reviewing MDE CopilotKit and AG-UI changes in a pull request.