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/kimmingul/xllm/claude-mdgit clone --depth 1 https://github.com/kimmingul/xllmWrote 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/kimmingul/xllm/claude-md)<a href="https://agentmods.dev/instructions/kimmingul/xllm/claude-md"><img src="https://agentmods.dev/badge/instructions/kimmingul/xllm/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 | $0.00553 | $0.00553 |
| Opus 5 | $0.00277 | $0.00277 |
| Sonnet 5 | $0.00111 | $0.00111 |
| Haiku 4.5 | $0.00055 | $0.00055 |
Grade A, and why
xllm 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 5d 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.
What it actually says
CLAUDE.md — xllm
먼저 읽기: CONTRIBUTING.md(개발 루프·릴리스 체크리스트·관례) · docs/SCOPE.md(만들지 않을 것).
명령
npm run ci— check+test+smoke+bench selftest. 라이브 LLM 없이 통과해야 하며 GitHub Actions와 동일 구성npm test— 순수 함수 + 픽스처만(디스크/라이브 금지). 파괴적 동작 테스트는 격리가 실제로 작동했는지 단언할 것(XLLM_STATE_DIR 패턴)- 릴리스 = CONTRIBUTING 체크리스트 + 주석 태그
vX.Y.Z+gh release create(CHANGELOG 발췌를 노트로)
구조
- 어드바이저 진입점은
scripts/xllm-advisor.js하나 (grok-ask-advisor.js shim은 v0.21.0에서 제거됨) - 런타임 상태는
.xllm/(panel-ledger.jsonl·artifacts·xllm-providers.toml);.grok/은 Grok Build 어댑터 전용 - 증거는 append-only, 파생 뷰(stats/traits)는 절대 역기록 금지, 손으로 쓴 모델 인상론 금지(실측+표본 수 노출)
관례
- 비자명 설계는 codex@high×grok@high 적대 리뷰 2–3라운드(마지막 라운드는 검증된 코드 사실로 앵커) → docs/-design.md에 수렴 기록, 잔여 쟁점은 사용자 판정
- 프로토콜/전송 변경은 테스트만으로 출하 금지 — 라이브 e2e 후 출하
함정 (실측으로 진단됨)
- Windows argv ~32KB: 긴 프롬프트는
--prompt-file(검토 계열은 24KB 초과 시 자동 파일 경유); grok/gemini는 argv 전달이라 회피 불가, codex/claude는 stdin - ollama는 HTTP API(localhost:11434)로 호출;
ollama list/stop은 CLI 유지.ollama ps가 비었는데 cudaMalloc OOM이면 서버가 wedge된 것 — 재시작이 해법 - 양쪽 모델이 같은 단계에서 두 번 연속 실패하면 모델이 아니라 공유 전송층을 의심할 것
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.
- 5d ago First seen · 27 lines · 553 tokens per session scan A c310b5e05909
xllm CLAUDE.md is an instructions file published in the GitHub repository kimmingul/xllm (2 stars, last pushed 13d ago), licensed MIT. It adds 553 tokens to every session, about $0.0028 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 2026-08-31.
Other instructions, from other repositories
deploygate-agent-plugin CLAUDE.md
Claude Code instructions for DeployGate/deploygate-agent-plugin, covering claude.md, project overview, commands, architecture and testing.
ai-plugin GEMINI.md
Gemini CLI instructions for PostHog/ai-plugin, covering posthog extension, available tool categories and guidelines.
TokenDiet copilot-instructions.md
Instructions for Fortytude/TokenDiet: When you reason about LLM / token cost in this repo, apply the /tokendiet discipline — describe technical waste, never fabricate the dollars.
m2ui-skill copilot-instructions.md
Copilot instructions for martysama0134/m2ui-skill, covering m2ui — metin2 ui code generator, mode detection, reference documentation, two ui styles and critical code generation rules.
quantcoder-plugin GEMINI.md
Gemini CLI instructions for nutdnuy/quantcoder-plugin: Follow AGENTS.md. Use skills/quantcoder-research/SKILL.md as the primary task guide for QuantCoder research-paper to QuantConnect workflows.
ControlFlow copilot-instructions.md
Copilot instructions for Smithbox-ai/ControlFlow, covering controlflow agent system — shared policies, continuity, build and test, path resolution and failure classification.