메뉴
设计、改造、融合和验证 Agent skills，在需要梳理策略框架、重构 SKILL.md 结构、规划 `references/` 与 `scripts/` 分层、建立评测、优化 description 或排查触发问题时使用。适用于“帮我做一个 skill”、“重写或合并这个 skill”、“给 skill 做评测”、“为什么这个 skill 没触发”等请求。
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/dulltackle/kangkang-skills --skill creating-skill-pro명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
AI-driven presentation workflow for generating editable PPTX decks and slides, reconstructing page visuals, creating reusable Brand/Style/Layout/Deck workspaces, filling native PPTX templates, and enhancing finished PPTX files. Use when the user asks to create, generate, reconstruct, regenerate, beautify, redesign, template, fill, or enhance a presentation, PPT, PPTX, slide deck, or courseware — including adding narration or animation to one — requests a presentation-authored narrated/self-running video, or mentions ppt-master.
Implement a piece of work based on a spec or set of tickets.
Commit a completed ticket as one Conventional-Commit, verifying each acceptance criterion, then ticking it off on the ticket itself. Never closes the ticket and never pushes. Invoked by /implement at the end of a ticket session.
SOC 직업 분류 기준
SKILL.md 표시 중
| name | creating-skill-pro |
| description | 设计、改造、融合和验证 Agent skills，在需要梳理策略框架、重构 SKILL.md 结构、规划 `references/` 与 `scripts/` 分层、建立评测、优化 description 或排查触发问题时使用。适用于“帮我做一个 skill”、“重写或合并这个 skill”、“给 skill 做评测”、“为什么这个 skill 没触发”等请求。 |
本 skill 用于设计、改造和验证 Agent skill。默认先判断任务性质、产物和边界，再按需读取 references/ 或调用 scripts/，不默认把所有请求推进到完整闭环。
激发模型能力上限 = 策略哲学 + 最小完备工具集 + 必要事实说明
SKILL.md 只放定位、判断框架、边界和资源入口references/ 提供按需细节，不默认全读scripts/ 承接重复、确定性、易错的动作agents/、assets/、eval-viewer/ 是可选能力层，不是每次都要启用SKILL.md 只保留路由、原则、关键决策和入口说明references/scripts/eval、benchmark、assertion 等词README.md、CHANGELOG.md、安装说明description 必须同时说明：
描述要优先写用户意图和使用时机，再补常见说法、相关文件类型和相邻边界。
优先沿着这四个维度判断，而不是直接启动完整流程：
references/，哪些动作需要 scripts/开始前先识别当前请求属于哪一类：
references/authoring/patterns.md、references/authoring/output-patterns.md、references/authoring/example.mdreferences/authoring/workflows.mdscripts/init_skill.pyreferences/merging/merge-playbook.mdreferences/authoring/patterns.md、references/authoring/example.mdreferences/evaluation/workflow.md、references/evaluation/schemas.mdscripts/run_eval.py、scripts/run_loop.py、scripts/aggregate_benchmark.py、scripts/generate_report.pyagents/grader.md、agents/comparator.md、agents/analyzer.md、eval-viewer/generate_review.pyreferences/triggering/description-optimization.md、references/evaluation/schemas.mdscripts/improve_description.py、scripts/render_trigger_eval_review.py、scripts/run_eval.py不要默认把四类流程全部跑完。只执行当前任务真正需要的部分。
skill 默认采用三层结构：
name 和 description 等高频元数据，始终在上下文里SKILL.md：只保留哲学、判断框架、边界和资源入口references/、scripts/、assets/、evals/约定：
SKILL.md 实务上尽量控制在 500 行以内references/scripts/典型结构：
skill-name/
├── SKILL.md
├── references/
├── scripts/
├── assets/
└── evals/
在本 skill 中：
references/authoring/：顶层哲学、workflow 取舍、结构模式、示例references/merging/：融合策略、迁移收尾references/evaluation/：eval、grading、benchmark、viewerreferences/triggering/：description 与触发优化name、目录名、skill 定位一致description 同时覆盖“做什么 + 何时触发”SKILL.md 保持精简，且只承担哲学、判断和路由职责references/ / scripts/