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/lidge-jun/cli-jaw/agents-mdgit clone --depth 1 https://github.com/lidge-jun/cli-jawWhat 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 | $0.06770 | $0.06770 |
| Opus 5 | $0.03385 | $0.03385 |
| Sonnet 5 | $0.01354 | $0.01354 |
| Haiku 4.5 | $0.00677 | $0.00677 |
Grade C, and why
cli-jaw AGENTS.md scanned grade C with 2 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 2d 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
- If an old commit accidentally stages Rust build artifacts, stop before committing and run `git rm -r --cached src/rhwp-field-bridge/target` plus `rm -rf src/rhwp-field-bridge/target`; commit only source, fixtures, test Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- 반영 여부 검증: `grep <new-symbol> dist/...` 또는 해당 엔드포인트 curl로 확인 후 안내. How it starts
The opening of the file, as written. The whole thing — 359 lines — stays where its author put it; the contents beside it link to each section on GitHub.
cli-jaw
System-level AI agent with full computer control via CLI wrapping (pi, agy, ai-e, claude, claude-e, codex, codex-app, cursor, kiro-code, gemini, grok, opencode, copilot).
Repository Structure
lidge-jun/cli-jaw ← public (this repo)
├── skills_ref/ (submodule) ← lidge-jun/cli-jaw-skills (public reference skills)
├── devlog/ (submodule) ← lidge-jun/cli-jaw-internal (private)
└── .npmignore ← npm publish 시 submodules 제외
Remote / 브랜치 정책 (푸시 전 필수)
origin은 https://github.com/lidge-jun/cli-jaw.git 이다. 체크아웃에 따라
bitkyc08-arch/cli-jaw 로 남아 있을 수 있는데, 이건 같은 저장소의 옛 이름이라
GitHub이 리다이렉트해 준다. 동작은 하지만 gh(항상 lidge-jun/cli-jaw 로 해석)와
이름이 갈려서 "이슈는 A에, 푸시는 B에" 처럼 보이는 착시를 만든다. 발견하면 고칠 것:
git remote -v # 확인
git remote set-url origin https://github.com/lidge-jun/cli-jaw.git
작업은 전부 dev 에서 한다. main 과 preview 는 머지만 받는 브랜치다.
거기에 직접 커밋하지 않는다.
dev 는 항상 main 위에 fast-forward 가능한 상태로 유지한다. merge 커밋으로
main 을 끌어와 히스토리를 갈래지게 만들지 말고, rebase 로 main 바로 위에 올린다:
git fetch origin
git branch -f backup/dev-pre-rebase-$(date +%y%m%d) HEAD # 되돌릴 지점
git rebase origin/main
git merge-base --is-ancestor origin/main HEAD && echo ff-able
git push --force-with-lease origin dev
rebase 중 만나는 충돌은 대개 두 종류이고 해소법이 정해져 있다.
- 버전 파일(
package.json,package-lock.json,electron/*): main 쪽을 취한다 (--theirs). 이건 릴리스 장부이지 작업 산출물이 아니다. 오래된 preview 승격 커밋이 replay 되려 하면git rebase --skip— 버전을 되돌리는 것 말고 하는 일이 없다. structure/str_func.md: 어느 쪽도 고르지 말고 실제 트리에서 다시 만든다. 줄 수는 파생값이다 —bash structure/verify-counts.sh --fix후 재검증.
dev 푸시는 CI를 돌리지 않는다. test.yml 과 postinstall-platform.yml 은
preview/main push와 pull_request 에만 반응한다. exact-head CI 증거가
필요하면 PR 을 열어야 하며, dev 에 푸시해 두고 "CI 통과"를 주장하면 안 된다 —
애초에 아무것도 돌지 않는다. 확인은 SHA 대조로 한다:
git rev-parse HEAD
gh pr checks <n>
gh run view <run-id> --json headSha,conclusion
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.
- 2d ago First seen · 359 lines · 6,770 tokens per session scan C 9c87be0bbcbe
cli-jaw AGENTS.md is an instructions file published in the GitHub repository lidge-jun/cli-jaw (200 stars, last pushed 2d ago), licensed MIT. It adds 6,770 tokens to every session, about $0.0338 per session on Opus 5. A static security scan graded it C with 2 findings (recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
Gladys AGENTS.md
AGENTS.md instructions for GladysAssistant/Gladys, covering agents.md, cursor cloud specific instructions, running the app (dev mode), non-obvious caveats and feature specs (spec-first process).
nullclaw CLAUDE.md
Instructions for nullclaw/nullclaw, covering claude.md, mandatory reference, build & test commands, requires exactly zig 0.16.0 (verify: zig version) and build flags.
openclaw CLAUDE.md
Claude Code instructions for openclaw/openclaw, a project described as: Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞.
reverse-engineering-assistant copilot-instructions.md
Instructions for cyberkaida/reverse-engineering-assistant, a project described as: MCP server for reverse engineering tasks in Ghidra 👩💻.
openyida CLAUDE.md
Instructions for openyida/openyida, covering openyida — claude code 开发指引, 快速上手, 核心文件速查, 关键约定 and codex 特殊说明.
FERAL-AI CLAUDE.md
Instructions for FERAL-AI/FERAL-AI, covering claude.md — feral / asos, what this is, layout, commands and rather than restating it.