From agentic-sop-kit
Methodology to convert human SOPs into deterministic, gated agentic workflows with hard gates and human approval. Build/extend agent tools and design SOP→Skill→Workflow decomposition.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agentic-sop-kit:agentic-sopThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
- **用**:要把人工 SOP/手動流程自動化、建新 agent 工具、設計「拆解+閘門」、或開新的 agent 工作流專案時。任何專案皆適用,尤其未來新項目。
agentic-workflow-audit;執行既有 GMP 產生器 → 用該工具自己的 skill。Human SOP → 工具 Skill(SKILL.md) → Agentic Workflow(編排 + 誠實閘門 + hook)。 中間的 Skill 環最易被忘:工具一改,SKILL.md 必須同次更新。每支工具實作同一條七階段迴圈: intake → 分類/前置檢查 → 確定性層(程式) → 生成層(Claude,只整理輸入) → 組裝 DRAFT → 閘門自評(≤2 次) → 覆核包 → 人核准 STOP。
方法論已封裝成可攜套件,直接導入、別重寫:
python3 ~/.claude/agentic-sop-kit/bootstrap.py --project /path/to/project(複製 kit+裝 /sop-flow+合併 Stop-hook)。canonical 在 ~/.claude/agentic-sop-kit/。python3 agentic-sop-kit/check_deps.py 驗依賴;workflow/run.py 跑通範例。templates/human_sop_template.md 寫 SOP(每步標工具);依拆解規則把每個「步驟×工具」用 templates/skill_template/ 建成 skills/<name>/。workflow/flow.json 按 SOP 順序接線;裝 commands/ + hooks/。
詳見 kit 的 SOP.md(方法論 canonical)與 README.md(安裝)。本 skill 是提醒與指引(散文,機率性)。真正的強制力在 kit 的 Stop-hook 自動回歸:更新後有變動才跑「受影響單元+整合」兩層,fail 就 decision:block 把原因餵回去修,附重試上限防迴圈。要讓某專案「真的被擋」,就把 kit 的 hooks/(SessionStart 依賴檢查 + Stop 回歸)裝上。
npx claudepluginhub s0912758806p/agentic-sop-to-work --plugin agentic-sop-kitCreates, updates, and validates Agent SOPs — markdown-based workflows with RFC 2119 constraints for guiding AI agents through multi-step tasks.
Designs agent UX patterns and human-in-the-loop flows: autonomy levels (L0-L5), inbox pattern, progressive trust, decision journals, gate reviews. Six-phase methodology from pain point to data model.
Audits whether a project is a truly decomposed agentic workflow or a disguised mega-agent by examining task boundaries, I/O contracts, success criteria, SOPs, control flow, and failure rollback.