tapflow: Command for Claude Code

.claude/commands/ai-tells.md

ai-tells is a command for Claude Code from jo-duchan/tapflow. It costs 69 tokens per session (1,366 once invoked), scanned A, original, MIT.

A writing checker for spotting common signs of AI-generated prose in Korean or English, with optional manual rewriting or editing. It treats these signs as clues rather than proof that AI wrote the text.

In plain words
What is it for?
Use it to check documentation and other prose, or to review drafts before publishing; for Hacker News and Reddit posts, it supports checking but not automatic rewriting.
Why use it?
It gives people a way to review their own writing without treating an AI-detection score as reliable evidence of authorship.

Command for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: model in frontmatter.

This is jo-duchan/tapflow's own configuration. It tells Claude Code how to work on tapflow itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything tapflow configures →

Reuse

Borrowing it

Nothing to install: this file belongs to jo-duchan/tapflow. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/jo-duchan/tapflow/main/.claude/commands/ai-tells.md
Clone the repo
git clone --depth 1 https://github.com/jo-duchan/tapflow

Made for: Claude Code.

Wrote 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.

agentmods badge for ai-tells

README.md
[![agentmods](https://agentmods.dev/badge/commands/jo-duchan/tapflow/ai-tells.svg)](https://agentmods.dev/commands/jo-duchan/tapflow/ai-tells)
Your own site
<a href="https://agentmods.dev/commands/jo-duchan/tapflow/ai-tells"><img src="https://agentmods.dev/badge/commands/jo-duchan/tapflow/ai-tells.svg" alt="Measured on agentmods" height="20"></a>
Per session 69 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,366 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00069 $0.01366
Opus 5 $0.00034 $0.00683
Sonnet 5 $0.00014 $0.00273
Haiku 4.5 $0.00007 $0.00137

Measured 6d ago against content hash 575884c9625e, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

ai-tells 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 6d 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.

.claude/commands/ai-tells.md · 76 lines

How it starts

The opening of the file, as written. The whole thing — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.

입력: $ARGUMENTS — 형식 <lang> <mode> [target]

  • lang: ko | en
  • mode: detect | rewrite | edit
  • target: 점검할 텍스트, 또는 파일 경로(edit는 파일 필수). 생략 시 직전 대화에서 대상 글을 묻는다.

인자가 모호하면 멈추고 되묻는다. (예: /ai-tells en detect docs/guide/foo.md)


0. 이 도구의 정체 (먼저 읽는다)

  • 린트/게이트지 "세탁기"가 아니다. AI로 생성 → 톤 세탁 → 게시는 이 도구의 용도가 아니다. 사람이 쓴 글의 자가점검이 1차 용도다.
  • 신호지 증거가 아니다. 플래그된 패턴은 LLM 출력에 더 흔할 뿐, 비원어민·마감에 쫓긴 사람·기술 장르도 같은 모양을 만든다. 상용 AI 탐지기는 비원어민 영어에 오탐률 60%+. 점수를 "이건 AI다"의 단독 근거로 쓰지 않는다.
  • 운영 정책 (이 커맨드가 단일 소스) — 아래 규칙이 authoritative. 마케팅 채널별 상세는 로컬 .internal/marketing/OVERVIEW.md(비공개·gitignore, 클론엔 없을 수 있음)에 있고, 없으면 아래를 따른다.
    • detect = 기본. 커스텀 커맨드에서 자동 호출 가능.
    • rewrite = 수동. 자동 연결은 docs 경로(/write-docs 등) 한정. 마케팅·외부 포스트엔 자동 호출 금지.
    • HN·Reddit = 손작성. 외부 발견 채널 글은 detect 자가점검만, rewrite 쓰지 않는다.

1. 불가침 영역 (detect·rewrite·edit 공통, 절대 수정/플래그 제외)

다음은 탐지·교정 모두에서 건드리지 않는다:

  • 코드 블록, 인라인 코드, 명령어, 파일 경로
  • API·식별자·함수명·플래그·제품명·고유명사·기관명·모델명
  • 숫자·날짜·단위·버전·벤치마크 수치
  • 큰따옴표 안 직접 인용, 출처가 명시된 인용문
  • URL/링크 (단, utm_source=chatgpt.com 류 AI 도구 파라미터·citation 마크업 leak은 제거 대상 — rules 참조)

en rewrite는 산문(prose)만 손댄다. 위 항목이 섞인 문장은 해당 토막만 보존하고 산문 부분만 교정한다.

2. rewrite 가드 (정확성)

  • 변경률 30% 초과 = 경고, 50% 초과 = 중단·롤백. 원문 구조·의도·기술 디테일을 보존한다.
  • 원문에 없던 비유·수사·1인칭·의견을 임의로 추가하지 않는다.
  • register(격식체/평어체) 보존. 격식체 원문을 평어체로 떨어뜨리지 않는다.
  • 이미 사람다운 문단은 그대로 둔다. 플래그된 구간만 고친다.

3. 규칙 로드

  • lang == koRead .claude/ai-tells/rules-ko.md
  • lang == enRead .claude/ai-tells/rules-en.md

해당 파일의 패턴 분류표·심각도·처방을 기준으로 동작한다.

4. 모드별 동작·출력

detect (수정 없음)

  1. 발견: 패턴마다 해당 텍스트를 인용하고 ID·심각도(en: P0/P1/P2 · ko: S1/S2) 표기. 심각도순 그룹화.
  2. 판정: 각 플래그가 명백한 문제인지 의도적·맥락상 괜찮을 수 있는 것인지 구분. (잘 놓인 "however" 하나는 문제 아님.)
  3. 점수·등급: rules 파일의 등급 기준으로 자가 채점. 깨끗하면 깨끗하다고 말한다.

rewrite (산문 교정)

§1 불가침·§2 가드를 지키며:

  1. 발견: AI-ism 전부 인용.
  2. 교정본: 불가침 100% 보존한 정리본.
  3. 변경 요약: 의미 있는 수정만. 변경률 보고.
  4. 2차 점검: 교정본을 다시 읽어 살아남은 흔적(재활용 접속사·잔존 inflation·copula 회피 등) 잡고 인라인 수정.

Read the full file on GitHub · 76 lines

Changes

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.

  1. 6d ago Changed 575884c9625e
  2. 8d ago First seen · 76 lines · 69 tokens per session scan A 2885733a6337

Subscribe to this mod's changes

ai-tells is a command published in the GitHub repository jo-duchan/tapflow (591 stars, last pushed yesterday), licensed MIT. It adds 69 tokens to every session and 1,366 once invoked, about $0.0003 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.