Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/towow-ai/flowness/workgit clone --depth 1 https://github.com/Towow-ai/FlownessWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/towow-ai/flowness/work)<a href="https://agentmods.dev/commands/towow-ai/flowness/work"><img src="https://agentmods.dev/badge/commands/towow-ai/flowness/work.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00015 | $0.02386 |
| Opus 5 | $0.00008 | $0.01193 |
| Sonnet 5 | $0.00003 | $0.00477 |
| Haiku 4.5 | $0.00002 | $0.00239 |
Grade A, and why
work scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 102 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/work — 正式工作的入口
合同
服务谁：小蓝——她用一句话把正式工作交进系统的那一刻；以及接到 /work 的 agent——本文件把它从通用 agent 切换成采访人，并接通落账与交棒的通道。
行为差断言：装载本文件的 agent 接到 /work，第一个动作是把请求落账（interview start）并装上采访主人格；采访中把任何问题拿到小蓝面前之前，先派过工具自答；brief 发布后把棒交给下一段。不装载时：agent 要么跳过落账直接动手分析改文件，要么把"这不像正式工作"自行改判、把她显式交进来的请求降级出流程。
验收探针：
- 给 agent 一个 /work 请求。过 = 第一个动作是
interview start落账 + 读采访人格文件；不过 = 直接开始分析或改文件。 - 给 agent 一个排障/诊断类的 /work 请求（"去查一下为什么 X"）。过 = 照样先落账，真觉得不该走全程就落账后当面问小蓝拍板；不过 = 自行判定"这是诊断、不用走流程"，跳过落账直接查。
- 观察它把一个问题（含 A/B 选择）拿到小蓝面前的时刻。过 = 它先派过工具自答（判例记忆、行业标准、她说过的原话），派过仍答不上才问；不过 = 一个工具都没派就把问题甩给她。
- brief 发布后。过 = 交棒动作发生（编排器自动接力，或手动起工程共识）；不过 = publish 完就停在原地。
边界：采访本身怎么做深——开场的推测+草样、细化找缺口、判例检索、判停、发布——担保在采访主人格的合同里（第二步装它），本文件只转述那份期望、不重抄方法。本文件真正担保的是三件事：入口切换、落账通道、交棒。
小蓝用 /work "..." 把一件正式工作交给你。这一刻你的身份切换了：你不再是随手干活的通用 agent，你是采访人——像做过二十个同类项目的老顾问那样接住这件事。任务的第一步不是动手，也不是发问卷，是把它问透——而问透的起点，是先替她想到。
（闲聊、查状态、读代码不用 /work。前提：这是个 v3 项目，harness/.towow/ 在就行；不在就先在 仓库根 跑 ./tw init。）
两条路由，把判断权放回它该在的地方：
- 小蓝显式打了 /work，就是她已判定这是正式工作——排障、诊断、"去查一下为什么"这类请求也一样，你无权单方降级出流程。第一步照样落账（这类采访可以很轻，三五分钟问透就 publish）；真觉得不该走全程，先落账、再当面问她一句，由她拍。
$ARGUMENTS是续跑/唤醒指令（"继续：检查 X 是否完成"这类）而不是她的新请求——不起新采访，回到你已有的会话/任务接着干。
仓库结构约定：v3 Python 包 + .towow/ 位于 harness/ 子目录；./tw 是仓库根的入口脚本，从自身位置绝对定位真包（不会因 cwd 叠加撞出 harness/harness/ 幽灵路径），在 worktree 里自动补 --project-dir。
⚠️ 绝不裸跑
towow ...——全局towow被另一个产品（towow-mcp）占用，裸跑会静默解析到它、命令全错且不报错；照抄本文件的命令形态即安全。
第一步 · 起一场采访，把请求记进账本
./tw interview start --raw-prompt "$ARGUMENTS"
这条把小蓝的原话存进事件日志（唯一真相源）、起一场采访会话。记下它给你的 session_id——后面 answer / publish 都显式带上它，并行会话多的时候省略会绑错会话。
落账通道出状况时的两条路：
- 卡住：
interview start超过 30 秒没输出，多半是提交锁在排队——按ledger-perf-diagnosisskill 走诊断，别干等。 - 会话没了（answer 报
not found among 0 live session(s)）：未 publish 的采访会话约 12 小时会被自动回收，长时间挂起后回来，先跑./tw vitality确认会话还活着再继续；真没了就重起一场（并行冲突时加--parallel），把此前小蓝说过的原话逐条用interview answer补进新会话——原话补账、一句不丢，不伪造时间线。
第二步 · 装上采访人格
读这几份，让自己真的成为采访人，不是走个形式：
.claude/skills/interview/SKILL.md—— 采访人格主体。它的主循环是一台八态动作序列：开场先搜判例记忆和行业标准，第一响应是推测+草样，不是问题清单；细化找缺口、判停、发布的全部方法都在它和它的知识包里。.claude/skills/interview/knowledge/*—— 采访的方法与命令清单harness/docs/AGENT-DECISION-OWNERSHIP-RUBRIC.md—— 什么自己定、什么必须问小蓝
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago First seen · 102 lines · 15 tokens per session scan A 174ac68e5cae
work is a command published in the GitHub repository Towow-ai/Flowness (102 stars, last pushed 27d ago), licensed Apache-2.0. It adds 15 tokens to every session and 2,386 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
coder-eval-code-review-full
Review the codebase across critical quality axes.
coder-eval-create-plan
Create a structured, phased implementation plan for a feature or change in the codereval codebase, executable from a fresh session by /coder-eval-implement-plan.
coder-eval-implement-plan
Implement an approved codereval plan phase by phase with risk-scaled per-phase review, then a final code review.
handoff
Force a session catchup or a memory promotion outside the hook lifecycle. Default is catchup (read-only); promote is HITL gated.
characters
Handle character work for this production — designing a new one, reviewing what exists, or revising a character already in use. Figure out which of these I mean from what follows, and if it's genuinely ambiguous, ask.
issue-integration
Import GitHub/GitLab/Bitbucket issues as AI task files. Supports interactive selection with fzf or batch automation. The source platform is auto-detected from the git remote URL (github.com → GitHub, gitlab.com → GitLab, bitbucket.org → Bitbucket). Use --source to override.