Agent skill

Agentic Workflow Audit

by s0912758806p in s0912758806p/agentic-sop-to-work

A skill your agent uses when reviewing or auditing an existing agent / LLM-pipeline architecture — e.g.

MITAuto-check passedBusiness, Finance & HR

Install Agentic Workflow Audit

skills CLI
$ npx skills add s0912758806p/agentic-sop-to-work --skill agentic-workflow-audit -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install s0912758806p/agentic-sop-to-work agentic-workflow-audit --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/s0912758806p/agentic-sop-to-work.git skills-src && mkdir -p .claude/skills && cp -r skills-src/plugins/agentic-sop-kit/skills/agentic-workflow-audit .claude/skills/agentic-workflow-audit && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
agentic-workflow-audit
GitHub stars
209
Token cost
~1.2k tokens
SKILL.md length
236 words
Files
1
Skills in repo
11
Repo updated
First seen
Licence
MIT

At a glance

A skill your agent uses when reviewing or auditing an existing agent / LLM-pipeline architecture — e.g.

  • Works in 4 steps: 以程式碼與真實 trace / log 為準,不採信… → 每個判定都附證據:引用具體檔案路徑與行號,或一段真實 log / trace… → 不從寬解釋:模稜兩可時判 FAIL,並寫清楚你需要什麼證據才能改判。 → …
  • Auditing an existing agent / LLM-pipeline architecture — e.g
  • SKILL.md covers 角色與目標, 為什麼要這樣查, 行為準則 and 稽核項目, plus 3 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Agentic Workflow Audit is an agent skill from s0912758806p/agentic-sop-to-work. Use when reviewing or auditing an existing agent / LLM-pipeline architecture — e.g. 'is my workflow actually decomposed or secretly a mega-agent?', 'are my task boundaries and success criteria right?', 'is our shared state a blackboard?', 'can this retry loop run forever?' — even without the word 'audit'. | 要檢視/review/稽核既有 agent 或 LLM pipeline 的架構,或問「有沒有拆好」「是不是變成 mega agent」「task 邊界/成功標準對不對」「共享狀態是不是黑板」「退回會不會無限轉」「拓撲畫不畫得出來」時使用;任何評估既有 agent 系統結構的請求都觸發。不適用:要新建或自動化流程(改用 agentic-sop)。

Its SKILL.md is about 1.2k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Business, Finance & HR, covering Operations and SOPs. The repository describes itself as: Turn human SOPs into deterministic, gated agentic workflows — a Claude Code plugin. The licence is MIT.

When your agent uses it

  • Auditing an existing agent / LLM-pipeline architecture — e.g
  • Tasks that involve Operations and SOPs

Example prompts

  • “is my workflow actually decomposed or secretly a mega-agent?”
  • “are my task boundaries and success criteria right?”
  • “is our shared state a blackboard?”
  • “/agentic-workflow-audit”

Workflow steps

4 steps, taken from the first numbered list in SKILL.md.

  1. 以程式碼與真實 trace / log 為準,不採信 README、設計文件、註解裡的宣稱。
  2. 每個判定都附證據:引用具體檔案路徑與行號,或一段真實 log / trace 摘錄。無證據者一律標記 UNKNOWN。
  3. 不從寬解釋:模稜兩可時判 FAIL,並寫清楚你需要什麼證據才能改判。
  4. 找不到就標 UNKNOWN,絕不臆測為 PASS。

What it can do on your machine

Read from SKILL.md and the folder at commit 246d966. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Agentic Workflow Audit loads about 1.2k tokens when it runs. Until then it costs about 127 tokens; SKILL.md has 236 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~127
When it runs · the whole SKILL.md, loaded when a task matches
~1.2k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from s0912758806p/agentic-sop-to-work at commit 246d966, republished under its MIT licence (© s0912758806p). 236 words, ~1,154 tokens.

Download SKILL.mdSave it as .claude/skills/agentic-workflow-audit/SKILL.md (or your agent's skills folder).
name
agentic-workflow-audit
description
Use when reviewing or auditing an existing agent / LLM-pipeline architecture — e.g. 'is my workflow actually decomposed or secretly a mega-agent?', 'are my task boundaries and success criteria right?', 'is our shared state a blackboard?', 'can this retry loop run forever?' — even without the word 'audit'. | 要檢視/review/稽核既有 agent 或 LLM pipeline 的架構,或問「有沒有拆好」「是不是變成 mega agent」「task 邊界/成功標準對不對」「共享狀態是不是黑板」「退回會不會無限轉」「拓撲畫不畫得出來」時使用;任何評估既有 agent 系統結構的請求都觸發。不適用:要新建或自動化流程(改用 agentic-sop)。

Agentic Workflow 稽核

角色與目標

扮演一個唯讀的程式碼稽核者。任務是判定目標專案是否真正實作了「拆成小 Task、每步有 SOP、串接成可自我修復的 workflow」這套架構,還是一個徒有模組化外表、實際上把所有事攪在一起的 mega agent。

全程唯讀。不修改、不新增、不刪除任何檔案。

為什麼要這樣查

mega agent 的退化通常是悄悄發生的——程式碼看起來分了模組,跑起來其實全部黏在一起。文件與註解往往描述的是「意圖」而非「現況」。因此稽核的第一原則是看實際執行、不看宣稱。下面每一項檢查都要求你拿出證據,就是為了擋掉「自我安慰式」的從寬判定。

行為準則

  1. 以程式碼與真實 trace / log 為準,不採信 README、設計文件、註解裡的宣稱。
  2. 每個判定都附證據:引用具體檔案路徑與行號,或一段真實 log / trace 摘錄。無證據者一律標記 UNKNOWN。
  3. 不從寬解釋:模稜兩可時判 FAIL,並寫清楚你需要什麼證據才能改判。
  4. 找不到就標 UNKNOWN,絕不臆測為 PASS。

稽核項目

逐項執行下列七項。每項產出:判定(PASS / PARTIAL / FAIL / UNKNOWN)、證據、具體缺口、可執行的修補建議。

拿圖當檢查表。 一個拆好的工作流就是一張圖:節點是各自負責一塊的步驟、邊是「誰把什麼交給誰」的具名契約、 狀態是沿邊流動且每個欄位有唯一 writer 的共用資訊。下面七項就是在問這張圖畫不畫得出來、以及畫出來合不合法。 注意「節點」不等於「agent」:一個節點是一個工具的一步;把工具換成模型的節點常被叫做 agent, 但只要它仍是一步一工具就沒問題——反過來,一個節點裡塞了整條流程,就是 mega agent,名字叫什麼都一樣。

檢查 1 — 任務切分是否為真

能否在程式碼中明確框出每個 Task 的起點與終點。

  • PASS:每步有獨立、可定位的程式邊界,邏輯不與前後步驟混雜。
  • FAIL:步驟邏輯互相黏連,框不出單一步驟的範圍。
  • 試金石:能否將任一單一 Task 抽離、餵固定 input 獨立執行?無法在不啟動整條管線的情況下單跑某步 → FAIL。
檢查 2 — 步驟間是否有明確的 input / output 契約

步驟之間傳遞的資料是否有定義好的結構(schema / 型別 / 明確介面)。

  • PASS:每步輸入輸出結構明確且可驗證。
  • FAIL:所有步驟讀寫同一個大的共享狀態 / context,無誰給誰什麼的契約(黑板式共享狀態)。

分野:共用狀態本身不是問題,「無契約」才是。 上面那條 FAIL 的關鍵字是無誰給誰什麼的契約。 一份被具名邊與宣告式所有權約束的共用狀態,照這條規則寫法就是 PASS。判準是這三件事同時成立:

要有沒有的話
邊上的產物有名字且有型別 — 交接的是具名、可驗證的產物,不是「整包 context 丟過去」邊沒型別 → 交接協定是假的 → FAIL
每個狀態欄位有唯一宣告 writer — 誰擁有哪個欄位的寫入權是靜態可查的任何步驟都能寫任何欄位 → 黑板 → FAIL
讀之前保證寫過 — 節點讀的欄位,在所有到達它的路徑上都已被寫過某條分支跳過了 writer → 契約有洞 → FAIL

驗證方式:要求對方指出宣告在哪(哪個檔案、哪一行說了 owner 與型別)。 說不出來、只能說「大家都讀那個 dict」→ FAIL。若有靜態檢查器能在不執行的情況下判掉這三項 → PASS 的最強證據。 (agentic-sop-kit 的做法:reads/writes/schema_ref 宣告在 flow.json,lib/graph.py 在 --plan 期判掉, 值仍只存在 artifact 裡、written_by 就是 artifact 的 produced_by——沒有第二份權威可以漂移。)

檢查 3 — 每步是否有明確且可程式化檢查的成功標準

步驟跑完後,是否有程式碼明確判定「這次是否成功」。這是最常被偷工、卻最該嚴查的一項,因為它是回退自我修復能否運作的前提。

  • PASS:每步結束後有可程式化的成功條件檢查,並依結果決定推進或回退。
  • FAIL:做完直接呼叫下一步而無驗證;或「成功」僅等於「沒丟出例外」。
檢查 4 — 每步是否有獨立 SOP,且未被融進單一巨型 prompt

各步驟的作業規範是否各自獨立可見(獨立 prompt 檔 / SKILL.md / 文件)。

  • PASS:每步規範彼此分離、可單獨定位。
  • FAIL:存在一個包山包海的巨型 system prompt 把所有步驟規則全塞在一起。這是 mega agent 偷渡回來的最常見徵兆。
檢查 5 — 控制流由誰掌握

「下一步做什麼」由編排層程式決定,還是每輪交給模型自由決定。

  • PASS:流程走向可從編排碼直接讀懂(預定義路徑)。
  • FAIL:流程必須實際跑起來才知道模型會怎麼走(控制流落在單一模型手上)。
檢查 6 — 失敗邊是否存在,且是否有界

把「失敗處理」當成拓撲問題來查:圖上有沒有那條往回走的邊,以及那條邊會不會轉不停。

  • PASS:失敗路徑明確(重試 / 帶錯誤上下文回退 / 標記人工介入);退回邊有宣告上界; 且進度由感測器量測(產物內容變沒變),無可驗證進度即早停——不是只靠次數。
  • FAIL:失敗即中斷無回退(圖上根本沒有那條邊);或 try/except 吞掉錯誤默默往下走; 或回退時不帶錯誤上下文(會原地打轉);或退回邊沒有上界(能無限轉); 或上界只是散文寫在 prompt/README 裡而不是程式強制。
  • 逐一問清:哪條邊是退回邊?上界是多少?寫在哪一行程式?撞到上界之後會怎樣? 四題有一題答不出來 → FAIL。答「模型會自己判斷什麼時候停」→ FAIL(那是檢查 5 的問題)。
檢查 7 — 拓撲畫不畫得出來,且合不合法

不執行的情況下,能否從編排宣告畫出整張圖,並靜態判掉這些結構錯誤。

  • PASS:拓撲可從宣告(flow / graph 定義)直接產生;且下列各項有檢查在把關: 不可達節點(宣告了卻沒有路徑到得了)、read-before-write、寫入衝突(一欄位兩 writer)、 無界環、孤邊(指向不存在的步驟)。最強證據是這些檢查會讓 CI/dry-run 以非零退出碼失敗。
  • PARTIAL:畫得出來,但合法性只靠人看、沒有檢查。
  • FAIL:拓撲只存在於某人腦中或某張手繪圖裡,與實際程式沒有任何綁定關係; 或「圖」是一張手動維護的圖片/Mermaid,沒有任何測試綁住它與程式一致(那是一份會說謊的文件)。

三項決定性試金石(務必各自單獨執行並回報)

  1. 單步隔離執行:能否抽出任一步驟、以固定輸入單獨執行並驗證輸出?不能 → 切分不是真的。
  2. 僅憑 log 重建:只看一次真實執行的 log,能否清楚說出跑了哪幾步、每步輸入輸出、判定成功或失敗、是否觸發回退?不能 → 執行期沒有真正的任務分離,無論程式碼多模組化。
  3. 宣告畫圖、對照實走:只看編排宣告(不執行)畫出拓撲,再拿一次真實執行的紀錄比對—— 實際走過的節點順序(含重訪)是否都在那張圖的邊上?
    • 走過圖上沒有的邊 → 真正的控制流不在宣告裡(回頭看檢查 5)。
    • 紀錄裡看不出走了哪條邊、某節點被重訪幾次 → 觀測不足以稽核,判 UNKNOWN 而非 PASS。
    • 圖畫不出來 → 檢查 7 直接 FAIL。

可觀測性騙不了人,它直接反映底層結構——這三項是最能戳破「看起來模組化、其實攪在一起」的測試。 第 3 項尤其能抓到「文件上是一張漂亮的圖、跑起來是另一回事」。

輸出格式

務必照此結構回報:

## 逐項結果
(檢查 1–7,各列:判定 / 證據[檔案:行號 或 log 摘錄] / 具體缺口 / 修補建議)

## 拓撲
(用宣告畫出來的節點與邊;標出退回邊及其上界、每個狀態欄位的 owner;畫不出來就說明卡在哪)

## 試金石結果
(三項試金石各自的結論與依據)

## 總體判定(三選一)
- 真・拆解式 workflow:七項多數 PASS,三項試金石皆通過
- 部分退化:模組化存在但若干關鍵項 FAIL(點名是哪幾項)
- mega agent 傾向:控制流落在模型手上、或巨型 prompt 主導、或無法單步隔離、或拓撲只存在於腦中

## 最高風險項
(最該優先修補的 1–3 點,依嚴重度排序)

紅線

  • 不得修改任何檔案。
  • 每個判定必附證據;無證據即 UNKNOWN,不得臆測為 PASS。
  • 不得因文件或註解的宣稱而從寬判定。一張沒有測試綁住的架構圖也是宣稱,不是證據。
  • 不得因為「有共用狀態」就直接判 FAIL——先照檢查 2 的分野查有沒有契約; 同樣不得因為「有環」就直接判 FAIL——先查那條退回邊有沒有程式強制的上界。 從嚴的方向是要求證據,不是禁止設計。

© s0912758806p, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in plugins/agentic-sop-kit/skills/agentic-workflow-audit of s0912758806p/agentic-sop-to-work.

Open the folder on GitHubat commit 246d966

Compare with similar skills

Agentic Workflow Audit next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Agentic Workflow Audit compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Agentic Workflow Audit this skills0912758806p/agentic-sop-to-work209—~1.2kAutomated safety check: PassMIT
Cc Sdd New Agentgotalab/cc-sdd3.7k—~1.1kAutomated safety check: PassMIT
DBS Business Toolkit Entrydontbesilent2025/dbskill11k—~2kAutomated safety check: PassCustom licence
Agent Sop Authorstrands-agents/agent-sop1.2k—~3.5kAutomated safety check: PassApache-2.0
Diffusion Narrative Denouncingcanwhite/Krebs1k—~831Automated safety check: PassMIT
Polanyi Perspective0xenzyme/polanyi-skill137—~1.3kAutomated safety check: PassMIT

Similar skills

  • Cc Sdd New Agent

    gotalab/cc-sdd

    Add or extend coding-agent support in cc-sdd by executing the SOP in docs/cc-sdd/sop-new-agent.md end-to-end.

    3.7k GitHub stars~1.1k tokensUpdated 16 days ago
    Business, Finance & HRAuto-check passed
  • DBS Business Toolkit Entry

    dontbesilent2025/dbskill

    Chinese-language entry skill for the dontbesilent business toolkit: onboards new users, orchestrates tasks across sub-skills, runs numbered prompts and lists hidden ones.

    11k GitHub stars~2k tokensUpdated yesterday
    Business, Finance & HRAuto-check passed
  • Agent Sop Author

    strands-agents/agent-sop

    Create (or update) and validate Agent SOPs (Standard Operating Procedures) - markdown-based workflows that guide AI agents through complex, multi-step tasks with RFC 2119 constraints.

    1.2k GitHub stars~3.5k tokensUpdated yesterday
    Business, Finance & HRAuto-check passed
  • 基于"扩散模型叙事去噪流"的小说写作 SOP。将 AI 视为去杂质机器,通过锁定全局信号、预测叙事噪声、精准去噪、随机修正四个步骤,解决 AI 翻译腔、逻辑断层和故事平淡的问题。

    1k GitHub stars~831 tokensUpdated 1 mo ago
    Business, Finance & HRAuto-check passed
  • Polanyi Perspective

    0xenzyme/polanyi-skill

    Michael Polanyi 的思维框架。用 Polanyi 视角分析隐性知识、技能习得、经验传承、师徒制、 知识管理、学习方法、AI/工具替代边界、科学共同体与后批判哲学问题。

    137 GitHub stars~1.3k tokensUpdated 3 mo ago
    Business, Finance & HRAuto-check passed

More from s0912758806p/agentic-sop-to-work

All 11 skills in this repo
  • Agentic Sop

    s0912758806p/agentic-sop-to-work

    A skill your agent uses when turning a human SOP or manual process into an agentic workflow, building or adding an agent tool, or starting a new agent-workflow project — even if the user never says…

    209 GitHub stars~863 tokensUpdated 1 mo ago
    Auto-check passed
  • Six Rung Ladder

    s0912758806p/agentic-sop-to-work

    A skill your agent uses when deciding whether to write a piece of code, add a dependency, or introduce an abstraction / module / config — the "should I build this at all?" filter, applied before…

    209 GitHub stars~563 tokensUpdated 1 mo ago
    Auto-check passed
  • Alcoa Guard

    s0912758806p/agentic-sop-to-work

    A skill your agent uses when checking a record / dataset / GMP document for ALCOA+ data integrity — attribution, contemporaneity (backdating), completeness, accuracy (recompute / in-spec)…

    209 GitHub stars~363 tokensUpdated 1 mo ago
    Auto-check passed
  • Compute

    s0912758806p/agentic-sop-to-work

    A skill your agent uses to compute summary statistics (count/sum/mean/min/max) from a readings artifact, preserving the upstream source trace.

    209 GitHub stars~209 tokensUpdated 1 mo ago
    Auto-check passed
  • Draft

    s0912758806p/agentic-sop-to-work

    A skill your agent uses to produce a first draft, or to revise one from an upstream verdict artifact's open-defect count, emitting a draft@1 artifact.

    209 GitHub stars~250 tokensUpdated 1 mo ago
    Auto-check passed
  • Extract

    s0912758806p/agentic-sop-to-work

    A skill your agent uses to extract numeric readings (key: value lines) from a plain-text input into a structured readings artifact with source-location trace.

    209 GitHub stars~229 tokensUpdated 1 mo ago
    Auto-check passed

Questions about Agentic Workflow Audit

What does Agentic Workflow Audit do?

A skill your agent uses when reviewing or auditing an existing agent / LLM-pipeline architecture — e.g. Agentic Workflow Audit is an agent skill from s0912758806p/agentic-sop-to-work.g.

When should I use Agentic Workflow Audit?

Agentic Workflow Audit fits situations like: auditing an existing agent / LLM-pipeline architecture — e.g; tasks that involve Operations and SOPs.

How do I install Agentic Workflow Audit in Claude Code?

Run `npx skills add s0912758806p/agentic-sop-to-work --skill agentic-workflow-audit -a claude-code`. Or copy the skill folder (plugins/agentic-sop-kit/skills/agentic-workflow-audit in s0912758806p/agentic-sop-to-work) into .claude/skills/agentic-workflow-audit in your project. Claude Code loads it when a task matches its description.

How do I install Agentic Workflow Audit in Codex?

Run `npx skills add s0912758806p/agentic-sop-to-work --skill agentic-workflow-audit -a codex`. Or copy the skill folder (plugins/agentic-sop-kit/skills/agentic-workflow-audit in s0912758806p/agentic-sop-to-work) into .agents/skills/agentic-workflow-audit in your project. Codex loads it when a task matches its description.

Can I use Agentic Workflow Audit in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add s0912758806p/agentic-sop-to-work --skill agentic-workflow-audit -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/agentic-workflow-audit, .gemini/skills/agentic-workflow-audit, .github/skills/agentic-workflow-audit and .opencode/skills/agentic-workflow-audit in your project.

What does Agentic Workflow Audit need to run?

SKILL.md names no scripts, command-line tools or credentials: Agentic Workflow Audit is instructions for the agent only.

Does Agentic Workflow Audit access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Agentic Workflow Audit safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Agentic Workflow Audit use?

Agentic Workflow Audit is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Agentic Workflow Audit use?

About 1.2k tokens (SKILL.md is roughly 4.6k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Agentic Workflow Audit?

Skills that share tags, products or a category with Agentic Workflow Audit: Cc Sdd New Agent (gotalab/cc-sdd, 3.7k stars), DBS Business Toolkit Entry (dontbesilent2025/dbskill, 11k stars), Agent Sop Author (strands-agents/agent-sop, 1.2k stars) and Diffusion Narrative Denouncing (canwhite/Krebs, 1k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Agentic Workflow Audit?

s0912758806p (a GitHub user) maintains it in s0912758806p/agentic-sop-to-work, which has 209 GitHub stars. The repository holds 11 skills in this directory. The repository was last updated on September 8, 2026.

Source: s0912758806p/agentic-sop-to-work on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.