Skip to main content

workflow

将用户要求固化的已验证执行记录整理为可复用工作流或 Skill。

소스 정보

저장소
bronc-x/lotus
최근 소스 활동
2026년 9월 14일 05:31
감지된 SKILL.md 언어
영어
스타
23
포크
1

설치 방법

기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.

소스 파일 검토

설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.

파일 탐색기
18 개 파일

SKILL.md 표시 중

SKILL.md
소스 지침 · 읽기 전용 미리보기
name
workflow
description
将用户要求固化的已验证执行记录整理为可复用工作流或 Skill。
# Workflow Distill a real successful execution into a reusable path. Plans, prompts, and unverified drafts may produce a `DRAFT`, but they are not evidence of a proven workflow. ## Route - To reconstruct and simplify the executed path: read [references/distillation-method.md](references/distillation-method.md). - To create a workflow pack or Codex Skill: read [references/package-contract.md](references/package-contract.md). - To claim `VERIFIED`, irreducible, shortest, or optimal: read [references/validation.md](references/validation.md). Load only the references needed for the requested output and claim level. ## Invariants - Freeze the accepted result, scope, and evidence before optimizing the path. - Preserve correctness, permissions, safety, privacy, traceability, reproducibility, and diagnosable failure behavior. - Separate stable invariants from project defaults, conditional branches, local details, and detours. - Do not promote a one-off failure into a universal rule without repeatable evidence. - Removing steps can prove irreducibility only within the tested scope. “Shortest” requires comparison with other passing candidates; do not claim global optimality. - A workflow does not grant deployment, publication, deletion, production mutation, or other permissions absent from the original task. The bundled scripts can create and validate a workflow pack. Preview untrusted validation commands before execution; the runner is not a sandbox. Completion means a real success case is bound to the workflow, relevant negative cases and gates work, a fresh execution can follow the package, failures have been fixed and rechecked, and the evidence level is reported without overstating optimization.
GitHub에서 보기