Aliyun Skill Creator — 阿里云技能创建者

Category: 工具

Alibaba Cloud 技能 创建器

仓库-specific 技能 engineering 工作流 for alicloud-技能s.

Use this 技能 when Creating a new 技能 under 技能s/. 导入ing an external 技能 and adapting it to this 仓库. Updating 技能 trigger 质量 (name and description in frontmatter). 添加ing or fixing smoke tests under tests/. 运行ning structured benchmark loops before merge. Do not use this 技能 when The user only needs to 执行 an existing product 技能. The task is purely 应用 code under 应用s/ with no 技能 changes. 仓库 constrAInts (must enforce) 技能s live under 技能s///<技能-name>/. 技能 folder names use kebab-case and should 启动 with alicloud-. Every 技能 must include 技能.md frontmatter with name and description. 技能s//技能.md content must stay English-only. Smoke tests must be in tests///<技能-name>-test/技能.md. 生成d evidence goes to 输出/<技能-or-test-技能>/ only. If 技能 inventory changes, refresh README 索引 with scripts/更新_技能_索引.sh. Standard deliverable layout 技能s///<技能-name>/ ├── 技能.md ├── 代理s/openAI.yaml ├── references/ │ └── sources.md └── scripts/ (optional)

tests///<技能-name>-test/ └── 技能.md

工作流 Capture intent Confirm domAIn/subdomAIn and tar获取 技能 name. Confirm whether this is new creation, 迁移, or refactor. Confirm expected 输出s and 成功 criteria. Implement 技能 changes For new 技能s: scaffold structure and draft 技能.md + 代理s/openAI.yaml. For 迁移 from external repo: copy full source tree first, then adapt. Keep adaptation minimal but explicit: Replace 环境-specific instructions that do not match this repo. 添加 仓库 验证 and 输出 discipline sections. Keep reusable bundled resources (scripts/, references/, as设置s/). 添加 smoke test 创建 or 更新 tests//<技能-name>-test/技能.md. Keep it minimal, reproducible, and low-risk. Include exact pass criteria and evidence location. 验证 locally

运行 script compile 验证 for the 技能:

python3 tests/common/compile_技能_scripts.py \ --技能-path 技能s///<技能-name> \ --输出 输出/<技能-name>-test/compile-检查.json

Refresh 技能 索引 when inventory changed:

scripts/更新_技能_索引.sh

Confirm 索引 presence:

rg -n "<技能-name>" README.md README.zh-CN.md README.zh-TW.md

Optional broader 检查s:

make test make build-命令行工具

Benchmark loop (optional, for major 技能s)

If the user asks for quantitative 技能 evaluation, reuse bundled 工具ing:

scripts/运行_eval.py scripts/聚合_benchmark.py eval-viewer/生成_review.py

Prefer placing benchmark artifacts in a sibling workspace directory and keep per-iteration 输出s.

Definition of done 技能 path and naming follow 仓库 conventions. Frontmatter is complete and trigger description is explicit. Test 技能 exists and has objective pass criteria. 验证 artifacts are saved under 输出/. README 技能 索引 is refreshed if inventory changed. References references/模式s.md references/sources.md