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 agents/global-mindee/way/usage-guidegit clone --depth 1 https://github.com/Global-mindee/WAYWhat 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 | $0.00000 | $0.02591 |
| Opus 5 | $0.00000 | $0.01295 |
| Sonnet 5 | $0.00000 | $0.00518 |
| Haiku 4.5 | $0.00000 | $0.00259 |
Grade A, and why
USAGE-GUIDE 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 yesterday.
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 — 213 lines — stays where its author put it; the contents beside it link to each section on GitHub.
agents/USAGE-GUIDE.md — AI 강제 사용 지침서
목적: AI가 사용자 발화 → 적합 agent 1개 자동 선택·호출하는 결정 트리·프로토콜 버전: v1.1 (2026-06-11) 강제 로드:
CLAUDE.md 0.작업 시작 절차에 자동 로드 명시 (의무) 참조:agents/INDEX.md(활성 14 agent 압축 메타 + 네이티브 등록 loop-* 5종)
1. 강제 로드 선언
skills/USAGE-GUIDE.md 1. 과 동일 메커니즘. CLAUDE.md 0. 절차에 다음이 추가되어 있어야 함:
5. agents/USAGE-GUIDE.md + agents/INDEX.md (agent 선택 결정 트리)
2. 결정 트리 (5단계) — skills와 동일 패턴
Step 1. 사용자 발화 수신
의도 1~3문장 내부 CoT.
Step 2. agents/INDEX.md 의미론적 평가
4개 카테고리 절(활성 14 agent) 전수 후보 → description·trigger 의미론적 비교.
Step 3. 후보 분기
| 후보 수 | 처리 |
|---|---|
| 0개 | "관련 agent 없음" 명시 → skills 매칭 재시도 → 둘 다 0이면 일반 응답 |
| 1개 | 자동 채택 + 이유 1줄 알림 + 호출 |
| 2개 이상 | 점수 최고 1개 자동 채택 + 이유 1줄 알림 + 호출 |
Step 4. Critical Path 보강
rules/critical-path.md 4기준 해당 시 사용자 1줄 확인 + logs/operations.log 기록.
Step 5. agent 호출 + 본 작업 진행
- 호출 형식 참조.
3. agent 호출 형식 — 인라인 주입 패턴
Claude Code의 Agent 도구는 글로벌에 등록된 subagent_type 만 호출 가능 (예: general-purpose, Explore, Plan).
본 하네스 agents/는 로컬 정의이므로 그대로 subagent_type 으로 사용 불가. 대신 다음 패턴으로 인라인 주입:
Agent({
subagent_type: "general-purpose",
description: "<agent name 기반 짧은 설명>",
prompt: "
<agent-definition file=\"agents/01_dev-workflow/developer.md\">
{agent 본문 전체 인라인}
</agent-definition>
위 agent-definition을 따라 다음 작업을 수행:
{사용자의 원 요청}
"
})
이 방식은 본 plan(virtual-hopping-catmull) Phase 3 결재에서 채택. 향후 settings.json 정식 등록이 필요해지면 별도 결재 등록.
3-1. 도구 제약 frontmatter 표준 (C2 — 외부 하네스 클론 AgentDefinition 차용)
agent를 정의할 때 권한·라우팅 경계를 frontmatter(또는 본문 표)에 5속성 표준으로 명시한다. 검수자·실행자가 추측 없이 권한 경계를 알 수 있게 하기 위함이며, 특히 읽기 전용 비평 에이전트(critic)나 외부 영향 차단이 필요한 단계에서 의무다. 외부 하네스 클론의 AgentDefinition 속성 체계를 본 하네스 관행으로 번역해 채용.
| 속성 | 의미 | 값 예시 |
|---|---|---|
| tools | 허용 도구 화이트리스트 | Read, Grep, Glob, Bash(읽기 전용) — 쓰기·편집·커밋·push 금지 시 명시 |
| bashAllowedPrefixes | 허용 bash 명령 prefix(쓰기 차단의 핵심) | ls, cat, head, git log, git diff, grep, node --check |
| thinking-level | 사고 깊이(난이도 대비 비용) | low / medium / high |
| model-class | 모델 등급(품질 게이트) | frontier(최고) / standard / fast |
| routing-role | 파이프라인 내 역할·에스컬레이션 규칙 | specialist(전담, 범위 밖은 상위 위임) / orchestrator / worker |
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.
- yesterday First seen · 213 lines · 0 tokens per session scan A 5de3e80eb52a
USAGE-GUIDE is an agent published in the GitHub repository Global-mindee/WAY (11 stars, last pushed 1mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,591 tokens. 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 agents, from other repositories
Forge
OpenAI-family cross-vendor agent — runs OpenAI's flagship coding model via codex exec (ID resolved from CROSSVENDOR in models.ts, never pinned in prose). TWO MODES set by the spawn prompt. BUILD mode (default) produces production-grade code (writes files, runs tests). AUDIT mode (read-only) is the cross-vendor…
Max
Anthropic-family deep-analysis agent — the top-rung scrutiny pass added on top of super-sensitive work (public LifeOS releases, security boundaries, irreversible or publish-bound actions). Pinned to the fable alias, which always resolves to Anthropic's latest top model. Read-only: Edit/Write/NotebookEdit are denied at…
ClaudeResearcher
Academic researcher using Claude's WebSearch. Called BY Research skill workflows only. Excels at multi-query decomposition, parallel search execution, and synthesizing scholarly sources.
Grok
I am Jax. I run xAI's latest Grok model through LIFEOS/TOOLS/GrokQuery.ts, and I exist for exactly one lane: non-sensitive work on public data. A fourth vendor's distribution is genuinely useful — Grok has a different training signal (X-native, current-culture-heavy) than the Anthropic, OpenAI, and Google lanes — but…
Executive Orchestrator
Coordinate multi-workflow sessions spanning research, docs, scripting, and fleet changes — sequence executive agents and maintain session coherence.
Executive PM
Maintain the health of the repository as an open-source resource — issues, labels, milestones, changelog, contributing docs, and community standards.