Getting it into your agent
There is no command for this one: it runs only inside a plugin, and the catalogue could not identify which plugin ships it. The source is linked below.
Wrote 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/xiaojiou176-open/notebooklab/agents-md)<a href="https://agentmods.dev/instructions/xiaojiou176-open/notebooklab/agents-md"><img src="https://agentmods.dev/badge/instructions/xiaojiou176-open/notebooklab/agents-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 5.1 | $0.01656 | $0.01656 |
| Opus 5 | $0.00828 | $0.00828 |
| Sonnet 5 | $0.00331 | $0.00331 |
| Haiku 4.5 | $0.00166 | $0.00166 |
Grade A, and why
notebooklab AGENTS.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.
How it starts
The opening of the file, as written. The whole thing — 116 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Internal operator guidance for this fork. Not canonical public collaboration policy; use
README.md,CONTRIBUTING.md,SECURITY.md, andSUPPORT.mdfor public-facing authority.
Repository-level operator guide.
Purpose
- Route contributors and agents to the correct module entrypoint quickly.
- Keep runtime, documentation, and review rules small and stable.
- Preserve a minimal, open-source-friendly repository surface.
Kept navigation surfaces
- Root:
AGENTS.md,CLAUDE.md - Main modules:
apps/web,services/api,packages/core,tests
Nearest-first rule
Read order is fixed:
- nearest
CLAUDE.md - nearest
AGENTS.md - root
CLAUDE.md - root
AGENTS.md - a targeted public doc under
docs/
Search-before-write
Run these commands before adding rules, docs, or workflow policy:
rg -n "<keyword>" AGENTS.md CLAUDE.md README*.md docs config contracts tooling services packages apps tests ops evals mutants
rg --files -g 'AGENTS.md' -g 'CLAUDE.md' -g 'README*.md'
Record the commands plus at least one evidence token such as README.md:1 or services/api/main.py:1.
Non-negotiable repo rules
- Do not track
.agents/,.agent/,.codex/,.claude/,.runtime-cache/, log directories, or log files. - Current policy is stricter than earlier closeout phases: there are now zero tracked exceptions under
.agents/. - If
.agents/material is needed locally for planning or handoff, keep it untracked and repo-ignored only. - Keep tracked docs and navigation files English-only.
- Keep the public docs surface minimal.
- Route Python bootstrap commands through
tooling/scripts/runtime/run_uv_managed.shor the CI container wrapper. - When testing Git path visibility for tracked host artifacts, distinguish
tracked in the indexfromignored by rules; do not usegit status --untracked-files=allas a proxy for both.
Shared Workstation Hygiene
- Treat browser instances, browser profiles, tabs, Docker containers, and cache directories as shared workstation resources, not disposable infinite capacity.
- Treat host-process control as a hard safety boundary:
- never add
killall,pkill, or pattern-based host cleanup to worker, test, CI, or local-dev paths - only stop repo-owned processes through recorded positive PID files under
.runtime-cache/local/pids/ - if a port is occupied by an unknown PID, treat it as a blocker instead of killing first and asking questions later
- desktop or browser cleanup must stay repo-owned and exact-match; broad host cleanup is out of scope
- never add
- Start with ownership-first machine awareness:
- identify which browser/profile/tab/container/cache surface belongs to this repo and this task
- treat any surface owned by another repo or another active L1 lane as off-limits unless the user explicitly authorizes intervention
- When a task needs Chrome, Chromium, Safari, Playwright, or another browser-backed surface:
- if more than six browser instances are already active on the workstation, do not open a new one until ownership is checked and a smaller-footprint path is ruled out
- open the minimum number of windows and tabs required
- prefer background execution; treat detached browser/runtime launch as review-required and only acceptable inside repo-owned browser roots or directly held child handles
- avoid stealing the current desktop focus unless foreground control is strictly required for the task
- close every browser window or tab you created when the task is done
- do not leave background browser processes running without a current task need
- when this repo provides a generated browser identity tab, keep it as the human-facing anchor for that lane when practical
- do not script Chrome private pin-tab, avatar, theme, or Dock customization as part of the normal repo bootstrap; keep those manual if desired
- Never reuse or take over browser instances, tabs, or profiles created by another repo or another L1 lane. Only use browser state you created for the current repo, or state the user explicitly assigned to this repo.
- If ownership is unclear, treat the browser or cache surface as off-limits and record the ambiguity instead of guessing.
- If a task requires a cloned or temporary browser profile, keep it isolated to the current repo/task and remove the temporary clone when finished. Do not accumulate abandoned browser-profile copies under cache roots.
- Login-state checks must stop early:
- use at most one or two clearly-owned browser/profile attempts for the current repo
- if those attempts show that human login or owner-owned session state is required, record an external blocker instead of opening more browser instances or profile clones
- Keep Docker/buildx/container state governed:
- do not leave throwaway containers, builders, or large temporary caches running after the task that created them is finished
- do not perform broad cleanup that could disrupt another active repo or another L1 lane without explicit user approval
- Host safety is hard policy:
- do not use
killall,pkill -f,kill -9,xargs kill,osascript, orSystem Eventsin tracked automation paths - do not signal negative/zero PIDs or broad port listeners that are not proven repo-owned
- local runtime stop/restart flows must target repo-recorded PID files or repo-owned command signatures only
- do not use
- Git / GitHub closeout must be equally disciplined:
- close or delete only those branches, worktrees, and PR tails whose unique work is already merged into
mainor otherwise proven superseded - treat dirty worktrees with unique local changes as live continuation lanes, not automatic cleanup targets
- close or delete only those branches, worktrees, and PR tails whose unique work is already merged into
- Current-repo Git / GitHub closeout writes are allowed only inside the user-authorized closeout scope.
- Default to read-only behavior on external accounts and platforms outside that scope. Do not publish releases, edit off-scope repository metadata, submit listings, change domain or DNS state, post externally, mutate settings, or take any other write action against the user's external accounts unless the user explicitly asked for that exact write in the current task.
- Repository-scoped Git and GitHub closeout writes require explicit task authorization and do not extend to release publication, homepage/metadata edits, domain/DNS actions, third-party listing submissions, social posting, or partnership/outreach writes.
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 · 116 lines · 1,656 tokens per session scan A a2a18d79a473
notebooklab AGENTS.md is an instructions file published in the GitHub repository xiaojiou176-open/notebooklab (1 stars, last pushed 12d ago), licensed MIT. It adds 1,656 tokens to every session, about $0.0083 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
chamber CLAUDE.md
Instructions for abm9111/chamber, covering working on chamber, the gates that matter are the probes, not the tests, verification rules that have each been paid for, traps in this codebase and law worth not breaking.
docsmint AGENTS.md
AGENTS.md instructions for HiAi-gg/docsmint, covering docsmint — agents.md, cheat-sheet — conventions, project documents, core and project-specific.
SoDam-WikiMate AGENTS.md
AGENTS.md instructions for sodam-ai/SoDam-WikiMate, covering wikimate 작업 규칙 (크로스툴 — claude code · codex · gemini 공용), 문체, 자동 동작 (자연어 트리거 — 강제 호출 불필요), 접근 규칙 (자동 감지 — 특정 도구에 한정하지 않음) and obsidian 규칙.
ariadne AGENTS.md
Instructions for pariyar07/ariadne, covering ariadne — agent instructions, what this repo contains, how skills are structured, routing and rules.
soia-open-skills AGENTS.md
AGENTS.md instructions for soia-team/soia-open-skills, covering agents.md - soia-open-skills, repository purpose, safety rules, validation and git workflow(本仓特有:默认分支是 main).
second-brain-starter-kit AGENTS.md
AGENTS.md instructions for tonydzi/second-brain-starter-kit, covering agents.md — working in this repo, what this repo is, layout, how to verify a change and conventions.