메뉴
This skill should be used when the user asks to "improve a prompt", "make this prompt better", "optimize this prompt", "refine this system prompt", or wants to apply concept elevation to compress and clarify LLM instructions.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Use when writing documentation, READMEs, explanations, blog posts, or any prose meant for humans to read
This skill should be used when the user asks to "rebase all open PRs", "update all PRs with latest main", "make sure every open PR is up to date with main", "sync all open PRs onto main", "rebase all my PRs", or wants to bring every open pull request current with its base branch by rebasing in parallel git worktrees, resolving conflicts with a team of agents, and force-pushing.
Groom and prioritize a project backlog — verify each candidate issue's claims against the current codebase, pick the most important N, rewrite them to a ready-to-work standard (Problem/Goals/Decisions/Outcomes/Verification), move them to Ready on the board, and close obsolete issues. Use when asked to "groom the backlog", "triage the backlog", "prioritize issues", "move issues to ready", or "what should we work on next".
This skill should be used when the user asks to "summarize a note", "summarize this article", "replace #summarize-later", "add a summary to a note", "generate takeaways", or mentions summarizing Obsidian notes. Replaces the #summarize-later tag with a high-level summary and actionable takeaways.
This skill should be used when the user asks to "update the changelog", "add changelog entries", "what changed since last release", or when updating CHANGELOG.md with user-facing changes. Follows Keep a Changelog format and filters out internal-only commits.
Use when writing, reviewing, or improving system prompts for AI tools, agents, chatbots, or LLM-powered products. Trigger phrases include "write a system prompt", "create a system prompt", "system prompt for X", "improve this prompt", "review my prompt", "prompt engineering".
| name | improve-prompt |
| user-invokable | true |
| argument-hint | <prompt text or file path containing the prompt to improve> |
| description | This skill should be used when the user asks to "improve a prompt", "make this prompt better", "optimize this prompt", "refine this system prompt", or wants to apply concept elevation to compress and clarify LLM instructions. |
Apply concept elevation to make prompts more concise, clear, and effective.
Concept elevation takes stock of disparate yet connected instructions in a prompt, then finds higher-level, clearer ways to express the sum of the ideas in a compressed form. This makes the LLM more adaptable to new situations instead of relying solely on specific examples or instructions.
Decompose: Break the prompt into its core goals and concepts. Identify what each section is trying to achieve.
Group: Organize related concepts into clusters. Look for instructions that serve the same underlying purpose.
Elevate: For each group, find the single higher-level idea that captures the sum of the group's instructions. Iterate on candidate idea-sums until the optimal compression is found.
Synthesize: Combine the elevated concepts into a final prompt. Verify nothing was lost, then identify and fix any remaining redundancy or vagueness.
Validate: Compare the improved prompt against the original to confirm it preserves all intent while being more concise.
Perform each step inside <well-named-xml-style> tags to make the reasoning auditable.
The improved prompt should be:
