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 instructions/ai-screams/hwpforge/claude-mdgit clone --depth 1 https://github.com/ai-screams/HwpForgeWrote 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/instructions/ai-screams/hwpforge/claude-md)<a href="https://agentmods.dev/instructions/ai-screams/hwpforge/claude-md"><img src="https://agentmods.dev/badge/instructions/ai-screams/hwpforge/claude-md.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 | $0.09318 | $0.09318 |
| Opus 5 | $0.04659 | $0.04659 |
| Sonnet 5 | $0.01864 | $0.01864 |
| Haiku 4.5 | $0.00932 | $0.00932 |
Grade A, and why
HwpForge CLAUDE.md 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 4d 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 — 458 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Project Overview
HwpForge is a Rust library for programmatic control of Korean HWP/HWPX document formats, designed with LLM-first principles. The goal is to enable AI agents (like Claude Code) to generate Korean government proposal documents using natural language + Markdown + YAML style templates.
Current Status (snapshot — 2026-08-25):
- HWPX codec: read/write shipped · Markdown bridge: read/write shipped
- HWP5 → HWPX converter path: active, style/layout fidelity line in progress
- CLI bindings: shipped · MCP bindings: shipped · Python bindings: stub
- Shared
tab/ordered·bullet·outline/ checkable-bullet semantics wired through core → blueprint → smithy. HWP5 checkable carries all three gotcha-#8 truth locations (bullet.checkedChar,bullet.paraHead.checkable,paraPr.checked). - Phase 12 HWP5→HWPX carry series (GSO shapes/equation/memo/dutmal/compose/indexmark/click-here·auto fields/cross-ref instId/document metadata/outline 1–10)
main머지 완료. - E6 IR 와이어-누출 상환 완료 (
0.9.0): Summery rename(A) · BookmarkName collapse(B) · raw wire 필드 제거(C) ·inst_id/SystemId→공유ObjectId(M2). H1(display_text)=Won't-do (ADR-009 §CLOSURE, memorye6-wire-leak-status-h1-wontdo.md). - 0.10.0 릴리스 (2026-07-02): colLine(다단 구분선) HWPX+HWP5 carry(breaking) + rmcp 2.0 보안(GHSA-89vp-x53w-74fx)·quick-xml 0.41 + 문서 정합.
- AI 편집(에이전트 편집) 전 에픽 배포 완료 (2026-07-12~24,
0.11.0→0.11.6): E1 누름틀 채우기(0.11.0,feat!:display_text의미 변경) · E2fill델타 API·fields(0.11.1) · E6 템플릿 스탬핑 W1(0.11.2) · E3 표 격자 주소+set-cell(0.11.3) · E6 W2 클래스-B 셀 스탬핑+layout_carrylinesegarray 보존(0.11.4) · E5 outline/read/diff 읽기 3표면(0.11.5) · E4 문단 구조 편집insert-para/delete-para(0.11.6, preserve-first 바이트 스플라이스 + reverse-delta self-verify). 설계/로드맵 =.docs/planning/2026-07-10-agent-editing-architecture.md(남은 후속: E4b 표 행 편집 ·<hp:t>줄 경계 분할 보존 등 backlog). - PDF Export 에픽 배포 (2026-07-26~08-12,
0.12.0→0.13.1): smithy-pdf 신설 — 조판 캐시 재생(계산 금지) 렌더: 표·머리말/꼬리말·쪽번호·폰트 파이프라인(face 축·fsType)·CLIto-pdf(스니핑·3채널 경고 DTO). 에픽 canonical =.docs/planning/2026-07-26-pdf-export-epic.md. - 이미지/글상자 렌더 에픽 W1b~W5 배포 (2026-08-14~25,
0.14.0→0.16.0): W1b 좌표 ledger(가시 textpos 통일,0.15.0) · W2 인라인 이미지(0.15.1) · W3 표 셀 이미지+축약점 earliest-preimage(0.15.2) · W4 글상자 렌더+ObjectPlacement 공용화(breaking — 도형 11종 hp:pos 캐리) · W5 글상자 내부 인라인+body 앵커 렌더+HWP5 앵커 비트 byte-ground →0.16.0(트리거 3막 사고 — Releasing 절 필수 조건 2건 참조). 에픽 canonical =.docs/planning/2026-08-13-image-textbox-epic.md(잔여: sub-line-height 인라인 이미지 · W6 마감 · CI 다이어트 제안.docs/planning/2026-08-25-ci-diet-proposal.md).
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.
- 4d ago First seen · 458 lines · 9,318 tokens per session scan A b582c41b12bb
HwpForge CLAUDE.md is an instructions file published in the GitHub repository ai-screams/HwpForge (18 stars, last pushed 3d ago), licensed Apache-2.0. It adds 9,318 tokens to every session, about $0.0466 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 instructions, from other repositories
kordoc CLAUDE.md
Claude Code instructions for chrisryugj/kordoc, covering claude.md, 프로젝트 개요, 빌드 & 개발, 코퍼스 동기화 (2026-08-22 기준선) and 기기 간 코퍼스 동기화 — rtk 훅이 rsync 를 리라이트해 출력을 삼키므로 절대경로로 호출.
mq copilot-instructions.md
Copilot instructions for harehare/mq, covering mq development guide, project overview, coding conventions, rust code conventions and commit message conventions.
mq commit.instructions.md
Instructions for harehare/mq: Use the following format for commit messages.
mq general.instructions.md
Instructions for harehare/mq, covering general coding conventions for mq, directory structure, feature requests and bug reports.
mq rust-crate.instructions.md
Instructions for harehare/mq, a project described as: A jq-like Markdown query language for command-line processing.
mq testing.instructions.md
Instructions for harehare/mq: Use just test-all to run tests instead of cargo test.