agentic-sop-kit
Convert human SOPs into deterministic agentic workflows. Bundles the six-rung-ladder + agentic-sop + agentic-workflow-audit skills, the /sop-flow command, a project-scoped Stop-hook regression gate, and the portable agentic-sop-kit.
pinned to #248a01fupdated 2 weeks ago
Ask your AI client: “install plugins/agentic-sop-kit”.
Requires the metahub MCP server installed in your client. Set up MCP.
mh install plugins/agentic-sop-kitmetahub onboarded this repo on the author's behalf.
If you own github.com/s0912758806p/agentic-sop-to-work on GitHub, claim the listing to take over publishing. Your claim preserves the existing eval history and badges; only the curator label is replaced with verified-publisher on your next publish.
Stars
197
Last commit
2 weeks ago
Latest release
published
- #agentic-workflow
- #claude-code
- #claude-code-plugin
- #claude-code-plugin-agenticsop
- #sop
Evaluation report
WarningsAutomated checks the publisher passed at publish time — structure, docs, safety, and whether the artifact behaves as claimed.248a01f· 2 weeks ago
Kind-specific
31Plugin: manifest fields complete
Plugin: bundled artifacts presentwarn
Couldn't find a skills/ subdir or an MCP server config inside the plugin
A plugin is most useful when it bundles ≥ 1 skill or an MCP server.
Plugin: bundle shape
empty bundle
Plugin: manifest location
manifest at plugin.json
Release history
1- releasecurrent248a01fwarn2 weeks ago
Contents
把「人工 SOP」轉成「確定性引擎 + 誠實硬閘門 + 人核准」的 agentic workflow。本 plugin 打包了方法論 skills、/sop-flow 指令、專案範圍的回歸 Stop-hook,以及可攜的 agentic-sop-kit。
內容
- Skills(model-invoked,安裝後到處可用)
agentic-sop— 把人工流程工程化成自動代理工作流的方法論與落地入口。agentic-workflow-audit— 稽核既有 workflow 是否真拆解,或退化成 mega agent。
- Command:
/agentic-sop-kit:sop-flow— 跑 kit 的 extract→compute→report 編排,回報 DRAFT。 - Hooks(專案範圍,關鍵設計)
SessionStart→ 若目前專案有agentic-sop-kit/就跑check_deps.py,否則略過。Stop→ 若目前專案有agentic-sop-kit/tests/verify.py就跑回歸閘門(變更偵測 → 兩層測試 → 失敗以decision:block餵回去修;重試上限防迴圈),否則略過。
kit/:可攜的agentic-sop-kit副本(含bootstrap.py、SOP.md、lib/、workflow/、tests/、templates/、範例 skills)。
引擎能力(kit 的 run.py 編排器,v1.5)
除了線性 demo,引擎還支援以下能力——全部確定性、由程式決定、附加式(既有流程照常運作):
- 每步硬閘門(產出後驗、fail 即停、零 LLM):
cmd_gate(exit 0)/schema_gate(必填欄位)/trace_gate(值須 verbatim 溯源 → 擋臆造)/recompute_gate(數字重算相符)。 cmd步驟(指令只來自flow.json白名單;會改動環境的需--allow-mutations)+--plan乾跑(執行前列出所有操作、標示 mutating,不執行)。- forward-only 條件分支(
branch/cases/goto)與map_over(對清單每項各跑一次、收集)——控制流由程式決定,不交給模型。 - 跨領域範例:
kit/workflow/examples/的免依賴 FE/BE/DB/AI 流程(各示範一個閘門)。 - 封頂自動修復 fix-loop:閘門失敗時
/sop-flow自動修復重跑(run.py --max-fix-retries,預設 3、程式強制上限),用盡才交人;永不竄改輸出過關(與 Stop-hook 回歸共用同一上限SOPKIT_MAX_FIX_RETRIES,預設 3——兩層、一個旋鈕)。 詳見kit/SOP.md(方法論)與kit/workflow/examples/README.md。
安裝(從 GitHub)
在 Claude Code(含 Claude Desktop 的 Code 分頁)執行:
/plugin marketplace add s0912758806p/agentic-sop-to-work
/plugin install agentic-sop-kit@agentic-sop-to-work
/reload-plugins # 或重開 session
也可用完整網址:
/plugin marketplace add https://github.com/s0912758806p/agentic-sop-to-work
驗證:/help 應看到 /agentic-sop-kit:sop-flow;兩支 skill 會依描述自動觸發。
開啟某專案的「強制力」
Hook 是專案範圍的:只有當該專案根目錄有 agentic-sop-kit/ 時,Stop 閘門才會驗證那個專案的 workflow(否則 no-op,不干擾無關專案)。要讓某專案受閘門保護:
python3 "<plugin 安裝路徑>/kit/bootstrap.py" --project /path/to/project
# 或用你機器上的 canonical kit:
python3 ~/.claude/agentic-sop-kit/bootstrap.py --project /path/to/project
之後在 Claude Code(含 Desktop Code 分頁)開該專案資料夾,Stop 時就會自動跑回歸閘門。
鐵則
- 事實只來自輸入,缺標【待補】,絕不臆造。
- 確定性的事用程式;硬閘門必須確定性,LLM 自評只能 advisory 且封頂。
- 產出一律 DRAFT + 人核准;受控/高風險判定永遠人擁有。
- 真正的強制力在 hook/gate(本 plugin 的
Stop閘門),不在散文。
方法論細節見 kit/SOP.md。
Reviews
No reviews yet. Be the first.
Related
mega-brain
OKF-powered knowledge context for Claude Code — injects your project knowledge base at every session
cc10x
The Loop Engine — harness and loop engineering plugin for Claude Code. Router-kernel orchestration, workflow artifacts, sub-agent delegation, hook enforcement, Test Honesty Gates, fresh-context verification, and memory that persists. 73% leaner than v11, zero quality regression.
manifest-dev-tools
Tools for working with prompts and PRs alongside the manifest workflow. Collaborative PR walkthroughs, autonomous PR review with per-comment verification, author-side PR babysitting via manifest-dev, gap-calibrated prompt engineering, cross-boundary context handoff, and incremental teaching for session work.
mh install plugins/agentic-sop-kit