LLM Mart Basic

가장 최근 윤문 결과를 2차로 다시 다듬는다 — 특정 카테고리·문단·강도 조정도 가능. humanize-korean strict 윤문(Phase B)을 기존 run_id에 재실행해 잔존 finding을 처리한다. 트리거 — "/humanize-redo".

AI(ChatGPT·Claude·Gemini)가 쓴 한글 텍스트를 사람이 쓴 글처럼 윤문한다. 번역투·영어 인용 과다·기계적 병렬·관용구·피동 남용·접속사 남발·리듬 균일·이모지/불릿 과다 등 10대 카테고리 70개 AI 티 패턴을 탐지·분류해 내용은 한 글자도 건드리지 않고 문체·리듬·표현만 자연스럽게 재작성한다. 트리거 — "AI 티 없애줘", "AI 윤문", "ChatGPT 티 제거", "번역투 고쳐", "사람이 쓴 것처럼", "humanize Korean". 단순 맞춤법 교정·번역·내용 추

정밀(strict) 모드 1단계 진단 에이전트. 글 전체를 한 번에 보고 "가장 지배적인 AI 티 패턴 3~6개"를 taxonomy ID와 함께 진단한다. 불안정한 span 열거(0↔18개로 요동) 대신 "무엇이 이 글을 지배하는가"라는 안정적 판단을 내려, 후속 윤문 콜이 그 진단을 겨냥하게 한다. 산출물은 02_diagnosis.md 1개. 도구 호출 3회 캡(Read 결합입력 + Read taxonomy + Write 진단). 이 진단이 정밀 모드 품질의 결정 변수다.

정밀(strict) 모드 3단계 마무리 에이전트. 원문과 윤문본을 직접 대조해 ①의미 보존(15항 — 각주·제목·없던 주장 주입 포함) ②자연성(잔존 AI 티 + 과윤문 양방향)을 한 콜로 병합 판정하고, 문제 구간만 국소 보정한다. 전체 재작성 금지 — 의미 드리프트(빈 수사를 없던 주장으로 대체)를 막는 게 존재 이유. 은퇴한 content-fidelity-auditor·naturalness-reviewer 2인을 대체한다. 산출물은 final.md + 09_finalize.json. 도구 호

v1.6.1 Fast Path 단일 호출 윤문 에이전트. 한 호출 안에서 탐지·윤문·자체검증을 일괄 수행하여 5,000자 이하 한글 입력을 2~3분 안에 처리한다. 산출물은 final.md 1개(본문 끝에 `<!-- HUMANIZE-SUMMARY -->` HTML 주석 블록으로 메트릭·등급·자체검증 통합). 도구 호출 chain 3회 캡. 깊은 검증이 필요하면 정밀 모드(진단→윤문→finalize 3콜) 사용.

AI가 생성한 한글 글의 "AI 티" 패턴을 체계적으로 분류·확장·버전 관리하는 도메인 전문가. `references/ai-tell-taxonomy.md`를 단일 진실 원천(SSOT)으로 유지하며, 실제 입력에서 관찰된 신규 패턴을 검증해 v1 → v2로 승격한다.

한국 번역학계(이근희·김정우·김도훈·김순영·김혜영·이영옥·곽은주·조의연)와 국제 번역학(Baker·Toury·Laviosa·Chesterman·Toral·Sarti)의 학술 인용 계보를 Humanize KR 본진 SSOT(taxonomy.md)와 외부 references/scholarship.md 양면에 안전하게 안착시키는 학술 정통성 큐레이터. 보고서의 학술 자산을 본진 분류 체계가 검증 가능한 형태로 흡수하되, SSOT 룰북 슬림성을 해치지 않게 메타필드 + 외부 파일로 분리. 본진 패턴에 출

Toral 2019 post-editese 3축(단순화·정규화·간섭)을 한국어 정량 지표로 구체화하고, 보고서 8유형 검출 시그널을 metrics.py에 추가해 회귀 검증 가능 상태로 만드는 정량 엔지니어. 표준 라이브러리만, 형태소 분석은 정규식·접미사 사전으로 근사(konlpy·mecab 금지 — v1.6 정책 보존). monolith 외부 사전 처리(prepare_monolith_input.py)에 결합되어 도구 호출 캡 3회 보존. 신규 metric 추가 또는 metric 회귀 검증 시 호

신규 분류 체계 v2.0과 metrics·playbook 패치를 quick-rules.md(monolith 전용 슬림 룰북, 126줄 → ≤180줄)에 안착하고 monolith 도구 호출 3회 캡(v1.6.1) 회귀를 검증한 뒤, GitHub PR 초안과 CHANGELOG를 작성하는 통합 엔지니어. 본진 룰북 슬림성·monolith 정의 무수정·v1.x 발행 정책(사용자 명시 승인 후 푸시)을 3대 가드로 삼음. v2.0 변경 묶음을 PR로 발행 직전 단계에서 호출.

Humanize KR 본진 v1.6 분류 체계(10대 카테고리·61+ 패턴)와 외부 학술 보고서 후보 풀(translationese-research-distiller 산출물)을 3-축 매트릭스(이미 본진·보강·신규)로 매핑해 분류학자에게 승격 결정 입력을 제공하는 갭 분석가. 사실 발견만 하고 승격 결정은 하지 않는다 — taxonomist가 최종 판정자. 본진 v1.6 → v2.0 업그레이드 회차 또는 외부 보고서를 본진과 합칠 때 호출.

한국어 번역투(translationese) 학술 보고서를 8유형·15항목 PE 체크리스트·post-editese 3축·학술 인용 계보·예문 코퍼스로 분해해 후속 분류·승격 단계가 직접 소비할 수 있는 구조화 JSON으로 증류하는 도메인 추출가. 보고서 본문에 명시된 사실만 추출하고 자체 추정·확장은 금지. 보고서가 한국 번역학계의 8대 번역투 유형(무생물 주어·피동·대명사·-들·관계절·have-make·조사 결합·종결어미)을 다루거나 Toral 2019 post-editese·Baker 1993

Handle ZenTao (禅道) Bugs and Tasks end to end, including updating or writing back an item after code changes, managing Task status and hours, and reading linked Stories. Use for referenced ZenTao items, requirements, status changes, time entries, or post-implementation synchroniza

Generate a Conventional Commits message from staged changes and wait for confirmation before committing. Use when the user asks to commit or generate a commit message.

Summarize each day's Git activity into a concise daily work log, for a single date or a range. Uses the current repository, optional configured work projects, or paths named in conversation; configuration is never required.

Review code changes for bugs, regressions, convention violations, and high-value cleanup opportunities. Use for diffs, commit ranges, hosted PR/MR URLs, branches, paths, staged changes, or working-tree changes.

Summarize a contributor's Git history, a provided work log, or both into a concise, review-friendly self-evaluation for quarterly, semi-annual, or promotion cycles.

Refactor changed code to reduce duplication, complexity, and wasted work. Use for diffs, commit ranges, PRs, paths, staged changes, or working-tree changes.

Query and display the current API provider balance and recent usage.

Inspect screenshots, photos, diagrams, image paths, and image URLs when the task depends on visible content. Use when the prompt lacks actual image content, native inspection fails, or the user requests inspect_image; prefer the MCP tool, then the installed CLI.

Use when working on accessibility, a11y, WCAG, ARIA, screen readers, keyboard nav, focus order, contrast, alt text, captions, reduced motion, or target sizes; not language/culture/device (see inclusive-design).

Fourteen posts of being wrong in production, compressed to checkboxes

Healthy nodes, a quiet network, 300 restarts in three days, and a latency budget measured in milliseconds

Discovery worked. Ping worked. Every TCP connection timed out, and later the tunnel only worked when someone had a terminal open.

Every VM came back. The cluster did not. Declarative systems converge on config, and the datapath isn't config.

A surprising share of AI-in-the-terminal failures aren't the AI. They're zsh, and a version of bash from 2006.

A Claude Code plugin turns standalone project configuration into a namespaced, installable extension that teams and communities can update as one unit.

None of the safety came from the model. It came from six boring habits.

Skills package instructions and references. Subagents run work in a separate context and return results. They solve different problems and can be composed deliberately.

Six hours in, one step left, everything green, and the incident that didn't happen

CLAUDE.md carries persistent project context. Skills load reusable procedures when relevant. Separating stable facts from task-specific workflows keeps both easier to maintain.

Twenty minutes recovering secrets that never existed, and the one sentence from a human that ended it

An API request routing a model's tool call through an approval gate to a remote MCP server

31 config keys, two audits, and why the first one was wrong in both directions

The official MCP Registry stores standardized server metadata rather than package code. Publishers verify a namespace, describe installation or remote access, and submit immutable versions.

Everyone looks at the Dockerfile. The file that actually leaked the key was the project file.

Remote MCP authorization uses established OAuth standards, but secure integration still requires issuer validation, least-privilege scopes, protected token handling, and server-side enforcement.

"Copy it over and switch the reference" is two steps, and the outage lives in the one nobody checks

stdio fits local processes and prototypes. Streamable HTTP fits hosted services and shared integrations. The right choice follows where the capability runs and who must reach it.

The most important rule wasn't about what I could change. It was about what I was allowed to display.

Tools perform operations, resources expose readable context, and prompts provide reusable templates. Choosing the correct primitive makes an MCP server easier to understand and govern.

List recent meetings and voice memos. Use when the user asks "what meetings did I have", "show my recent recordings", "any meetings today", "list my voice memos", or wants an overview of their meeting history. Also use when they need to find a specific meeting by browsing rather than searching.

Act as the user's live meeting sidekick inside the current terminal agent session. Use when the user explicitly asks you, the terminal agent, to watch a meeting, follow the live transcript, answer during the call, offer strategist thoughts, silently watch for risks, or track decisions. Do not use this skill to start or control the separate Minutes Coach HUD; explicit Coach or HUD lifecycle requests belong to minutes-copilot, and an ambiguous request such as "coach me live" requires one short surface clarification.

Self-coaching analysis of your own behavior across meetings — talk-time ratio, filler words, hedging language, monologue length, energy patterns, and (when meetings are tagged via /minutes-tag) what your behavior in winning meetings looks like vs losing ones. Use this whenever the user says "how did I do", "review my last meeting", "mirror", "self-review", "show my patterns", "coach me", "where am I weak", "talk time", "am I improving", "what do I do in meetings I win", "feedback on me", or asks for any kind of personal feedback on their own meeting behavior. This is the rare skill that gives the user a mirror to their own habits — surface it whenever they show curiosity about their own performance, even if they don't use the word "mirror".

Add a note to the current recording or annotate a past meeting. Use whenever the user says "note that", "remember this", "mark this as important", "add a note about", "annotate the meeting", or wants to capture a thought during or after a recording. Plain text input — no markdown needed.

Interactive meeting preparation — builds a relationship brief and talking points before a call. Use when the user says "prep me for my call with", "I'm meeting with X", "prepare me for", "what should I bring up with", "meeting prep", "get ready for my call", or wants to review history with someone before a meeting.

Generate a daily digest of today's policy-authorized meetings and voice memos — key decisions, action items, and themes across available recordings. Use when the user asks "recap my day", "what happened in my meetings today", "daily summary", "what did I discuss today", "any action items from today", or wants a consolidated view of the day's conversations.

Start or stop recording a meeting, call, or voice memo. Use this whenever the user says "record", "start recording", "capture this meeting", "stop recording", "I'm in a meeting", "take notes on this call", or wants to transcribe live audio. Also use when they ask about recording status or want to know if something is being recorded.

Draft user-facing Minutes release notes for a version from the commit range, recent GitHub releases, and the repository release checks. Use when the user asks to write, generate, prepare, revise, or review release notes or a changelog for a Minutes version.

Search past meeting transcripts and voice memos for specific topics, people, decisions, or ideas. Use this whenever the user asks "what did we discuss about X", "find that meeting where we talked about Y", "what did Alex say", "did we decide on", "what was that idea about", or any question that could be answered by searching their meeting history. Also use for "do I have any notes about" or "check my meetings for".

Guided first-time setup for Minutes — download whisper model, create directories, configure audio input. Use when the user says "set up minutes", "install minutes", "first time setup", "configure minutes", "get started with minutes", "how do I start using minutes", or when verify shows missing components.

Lightweight outcome tagging for meetings — won, lost, stalled, great, or noise. Use whenever the user says "tag this meeting", "mark that as a win", "that one was a loss", "tag yesterday's call as stalled", "mark this great", "that meeting was noise", "label that meeting", or any time they describe a meeting outcome in passing. Tagging takes 5 seconds and unlocks /minutes-mirror correlation analysis — the more meetings get tagged, the smarter mirror gets at telling the user what behavior patterns lead to wins. Surface this skill any time the user mentions a meeting result, win, loss, or wasted time.

Verify that Minutes is properly set up and working — model downloaded, mic accessible, directories exist, no stale state. Use when the user says "is minutes working", "check my setup", "verify minutes", "test recording setup", "why isn't minutes working", "minutes health check", or after running setup for the first time.

Analyze a product walkthrough, bug report video, Loom, or ScreenPal using Minutes transcription plus visual review. Use when the user wants a recorded demo or bug clip turned into a durable brief with transcript, key frames, issues, and next steps.

Weekly meeting synthesis — themes, decision arcs, stale commitments, and what deserves your attention next week. Use when the user says "weekly review", "what happened this week", "weekly summary", "recap my week", "any outstanding items", "week in review", or at the end of a work week.

Fast non-interactive briefing before any meeting — auto-detects your next calendar event, pulls relationship history, surfaces open commitments, and produces a one-page brief in under 30 seconds. Use this whenever the user says "brief me", "give me a quick brief", "what's coming up", "background on my next call", "who am I meeting next", "brief me on Sarah", "I have a call in 10 min", "quick rundown", or right before walking into a meeting. Different from /minutes-prep — brief is the fast hook-fireable version that doesn't ask questions and doesn't set goals. Use brief when speed matters; use prep when the user wants to think hard about goals first.

Manage old recordings — find large files, archive old meetings, delete processed originals. Use when the user says "clean up recordings", "how much space are meetings using", "delete old recordings", "archive meetings", "manage meeting storage", or asks about disk space from minutes.

Start and control Minutes Coach, the separate real-time copilot HUD, with an explicit meeting goal. Use only for explicit Coach or HUD lifecycle requests such as "start Minutes Coach", "open the Coach HUD", "pause Minutes Coach", "resume Minutes Coach", "Minutes Coach status", or "stop Minutes Coach". Do not use for requests that explicitly ask the current terminal agent to watch or strategize; those belong to minutes-live-sidekick. An ambiguous request such as "coach me live" requires one short surface clarification and must not automatically start Coach.

Post-meeting debrief — analyzes what happened, compares outcomes to your prep intentions, tracks decision evolution. Use when the user says "debrief", "what just happened in that meeting", "what did we decide", "debrief that call", "post-meeting", "what changed", or right after stopping a recording.

Policy-safe relationship rankings, commitments, aliases, person profiles, and topic research. Always use Minutes' bounded native CLI surfaces; never build or read a durable graph cache.

Surface recent voice memos and ideas captured from any device. Use when the user asks "what ideas did I have?", "what were my recent memos?", "what did I record while walking?", or wants to recall a captured thought.

Okou connects to the tools your team already uses and does the work — across marketing, sales, engineering, and operations, under your control.

An open-source Digital Worker platform for reliable execution, continuous co-evolution, and building Enterprise AI assets.

A curated list of AI Agent evolution, memory systems, multi-agent architectures, and self-improvement projects. | evomap.ai

A governance harness for AI coding.

DeepSeek Harness 手机版：可直接安装的 Android APK，AI 免 Root 操作手机（Shizuku/root 可选），文件编辑只需所有文件访问权限，前台保活 + AI 通知

Local-first, governed AI agent runtime for Python — embed it in your app, or run it as a CLI or ACP server. Permissions, MCP, memory and audit replay built in.

Orbi — the factory that builds and operates AI software factories. GitHub Issues in, releases and runnable system out

Open-source AI agent harness in native Rust — GUI, CLI, headless, and webapp from one binary. Multi-provider, MCP, skills, plugins, agent teams.

The SDK for browser agents. Interact, search, extract, and fetch any site reliably across the web

Agent-native shopping for extreme value: verifiable same-product price evidence, checkout, orders, delivery, and after-sales.

The extensible power-user platform for Google Antigravity. Adds a native In-App Browser, animated Desktop Pets, revamped Gemini UI, and custom BYOK Gemini Pro k…

Open-source, single-binary, self-hosted AI agent — your models and data stay on your machine. A coding agent on par with Claude Code and a personal assistant li…

Your pocket agent. Local-first AI agents on iOS and Android — real workspaces, tool execution with approvals, and your choice of model (DSH · Claude Code · Code…

Bridge between QQ (SnowLuma OneBot v11) and DeepSeek Harness agents: social simulation, safe MCP tools, slang learning and more.

A 7×6 framework for agent architecture. 28 patterns, each placed at a coordinate, runnable Python code with verified engineering slices from Claude Code, Aider,…

针对 AI 自动化渗透 Agent 的新一代反制蜜罐，通过反向代理将API密饵载入真实业务、反向提示词注入等方式反制自动化渗透 Agent，实现多款主流通用Agent的反制上线控制。

Open-source, local-first AI agent for coding and real work. BYOK models, MCP, skills, plugins, workflows, and private knowledge bases.

A free AI-agent toolbox for Android, 一站式安卓AI Agent工具箱

Google Workspace CLI — one command-line tool for Drive, Gmail, Calendar, Sheets, Docs, Chat, Admin, and more. Dynamically built from Google Discovery Service. I…

Independent, unofficial CLI to edit CapCut and JianYing (剪映) projects — subtitles, timing, speed, volume, templates, cut long-form to shorts. No API needed, rea…