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.
git clone --depth 1 https://github.com/gcake119/joplin-llm-wikiWrote 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/rules/gcake119/joplin-llm-wiki/joplin-brain-config)<a href="https://agentmods.dev/rules/gcake119/joplin-llm-wiki/joplin-brain-config"><img src="https://agentmods.dev/badge/rules/gcake119/joplin-llm-wiki/joplin-brain-config/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/rules/gcake119/joplin-llm-wiki/joplin-brain-config"><img src="https://agentmods.dev/badge/rules/gcake119/joplin-llm-wiki/joplin-brain-config.svg" alt="Reviewed on agentmods" width="80" 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 | $0.01684 | $0.01684 |
| Opus 5 | $0.00842 | $0.00842 |
| Sonnet 5 | $0.00337 | $0.00337 |
| Haiku 4.5 | $0.00168 | $0.00168 |
Grade A, and why
joplin-brain-config 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 11d 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
joplin-llm-wiki 設定與測試
- 知識流對齊：本 repo 部分採用
gatelynch/llm-knowledge-base四層概念。raw/對應raw/，wiki/對應wiki/，brainstorming/與artifacts/保持為探索與成品區；上游專案不是 runtime dependency。 - 語言：人可讀的知識管理輸出使用繁體中文；技術名詞、source path、filename 可保留原文。
- Joplin notebook 匯出：
sqlite-sync --select-notebooks寫入joplin_sqlite_sync.notebook_filter。來源路徑使用raw/<joined-notebook-slug>/<safe-title>.md；巢狀筆記本以-串接，例如工作/專案A/會議→工作-專案A-會議。reconcile_mode: mirror僅應清理匯出產生的 stale.md，不得刪非.md。 - SQLite sync 變更閘門：
sqlite-sync正常模式匯出後會比對 snapshot，只有 raw 變更才依joplin_sqlite_sync.pipeline.compile_mode觸發編譯。local觸發wiki-compile，agent觸發agent-compile，off不編譯。首次非 dry-run 只建立 baseline；--export-only匯出並更新 state 但不編譯；--snapshot-only只掃現有raw/建立 baseline，不開 SQLite、不刪檔、不編譯。舊run_wiki_compile僅作為未設定compile_mode時的相容 fallback。 - 定時語意：
sqlite-sync不是檔案系統 watcher。只有在sqlite-sync被執行時才會匯出 SQLite、比對 raw snapshot、依compile_mode觸發編譯。常駐輪詢需設定joplin_sqlite_sync.schedule.every_seconds或 CLI--every <seconds>；若用 launchd/cron 重複啟動單輪命令，保持every_seconds: null，避免外部排程與內部輪詢重疊。 - 觀測欄位：
sqlite-syncsummary 應維持raw_changed、change_detection、changed_files、compile_mode、compile_triggered，供 CLI、排程與 Health GUI 日誌判讀。 - Wiki 分層：
wiki-compile與agent-compile只能輸出到wiki/summaries/*.md、wiki/concepts/*.md、wiki/indexes/All-Sources.md、wiki/indexes/All-Concepts.md；summaries、concepts、indexes底下不得建立子資料夾。summaries是每個來源一份摘要，concepts是概念條目並交叉引用 summaries/concepts，indexes只放固定索引入口。source_refs使用raw下的相對路徑，例如工作-專案A-會議/my-note.md。 - Query / Capture：
query預設用--source-scope=knowledge，優先讀wiki/，必要時補raw/；--source-scope=wiki|raw可明確限制來源。成功回答不直接寫正式筆記，而是先建立 pending capture；確認後才寫brainstorming/chat/或artifacts/projects/<project>/。ask、index、watch、RAG／Chroma／embedding vector 管線已移除。 - 模型選擇：本地預設是
wiki-compile+ Ollama。Codex 月訂閱路線是agent-compile+ 本機已登入的codex exec，不使用 OpenAI API key，也不等同 API 額度。兩條管路預設都掃完整個raw/筆記庫；--batch=true才是 10-15 頁單批次 fallback。OpenAI API provider 目前未實作。 joplin_wiki_writeback.enabled預設為 true（YAML 省略該鍵時等同開啟）。當寫回視為開啟時，load-config要求joplin_data_api.token非空、joplin_data_api.base_url為合法 http(s) 且 hostname 僅允許127.0.0.1／localhost／::1，否則拋CONFIG_INVALID。- 在 單元／整合測試裡若只驗證
wiki-compile、query、sqlite-sync等、不需要真的呼叫 Joplin Data API，請在迷你設定中加入：
joplin_wiki_writeback:
enabled: false
若需測試 wiki 寫回路徑則設定 joplin_data_api.token（任意非空字串即可），並對 runWikiWriteback 傳入 fetch mock（見 test/joplin-wiki-writeback.test.js）。artifacts_project_notebook_title 只在 artifacts 按需寫回時必填。
wiki-compile --dry-run：不會對 Joplin 發送會變更資料的 HTTP；仍可能輸出writeback_would_write等乾跑統計（對 planner 路徑若檔案尚未存在，僅依路徑推估 topic）。corpus_auto_sweep.advance_state_on_dry_run: false（預設）時，dry-run 只跑 1 個 sweep 視窗且不推進corpus-sweep-state.json。- 小模型範例（見
config.yaml.example／README）：chat_model: gemma4:e4b、corpus_digest_max_files: 40、max_pages_per_run: 8、min_pages_per_run: 2、min_topic_pages_per_run: 3；wiki-compile預設單次掃完整輪，--batch=true才只跑一個本地批次；Ollamanum_ctx由 Modelfile 設定。 - 參考實作：
src/config/load-config.js、src/joplin/sqlite/exporter.js、src/joplin/sqlite/notebooks.js、src/joplin/sqlite/paths.js、src/joplin/sqlite/sync-state.js、src/commands/cmd-sqlite-sync.js、src/joplin/wiki-writeback.js、src/joplin/data-api-client.js、src/wiki/wiki-compiler.js、src/wiki/wiki-planner.js、src/commands/cmd-agent-compile.js。
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.
- 11d ago First seen · 29 lines · 1,684 tokens per session scan A f997dc265f1a
joplin-brain-config is a cursor rule published in the GitHub repository gcake119/joplin-llm-wiki (2 stars, last pushed 3mo ago), licensed MIT. It adds 1,684 tokens to every session, about $0.0084 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 cursor rules, from other repositories
agentmemory
AgentMemory MCP server rules — use addmemory before answering any question that introduces a project-specific convention, dependency, or constraint.
digital-brain
A Claude Code plugin and Obsidian vault that builds persistent, compounding knowledge bases using Andrej Karpathy's LLM Wiki pattern. This repo works with Cursor's AI through the cross-platform Agent Skills format.
obsidian-hub-spoke-graph
Hub-and-spoke (MOC) linking for Obsidian vault notes — no peer mesh, leaves up-link only.
vault-safety
Safety rules when writing to an Obsidian vault through Obsidian Bridge.
archcore-context
Archcore knowledge base context — document types, MCP tools, and conventions for working with .archcore/ documents.
transcreveai
Required handoff contract for nested TranscreveAI executions.