Borrowing it
Nothing to install: this file belongs to Ikalus1988/MisakaNet. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/Ikalus1988/MisakaNet/main/AGENTS.mdgit clone --depth 1 https://github.com/Ikalus1988/MisakaNetWrote 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/ikalus1988/misakanet/agents-md)<a href="https://agentmods.dev/instructions/ikalus1988/misakanet/agents-md"><img src="https://agentmods.dev/badge/instructions/ikalus1988/misakanet/agents-md/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/instructions/ikalus1988/misakanet/agents-md"><img src="https://agentmods.dev/badge/instructions/ikalus1988/misakanet/agents-md.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 | $0.05378 | $0.05378 |
| Opus 5.5 | $0.02151 | $0.02151 |
| Sonnet 5.5 | $0.01076 | $0.01076 |
| Haiku 4.5 | $0.00538 | $0.00538 |
Grade C, and why
MisakaNet 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 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
"params":{"name":"misakanet_preflight","arguments":{"intent":"rm -rf build/"}}}' Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
| **Streaming（SSE）** | `POST` + `Accept: application/json, text/event-stream` | 服务端以 `event: message` 分块返回；长任务/逐块消费用，`curl` 加 `-N` | How it starts
The opening of the file, as written. The whole thing — 238 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MisakaNet Agent 规则
本文件有两条主线，按你的角色选读： (A) 使用方 agent（来检索/贡献失败知识）→ §1–§5。 (B) 在本仓工作的 agent（改代码/写 lesson/维护 CI）→ §6 红线 +
docs/agents/repo-operations.md（完整工程手册）。 无论哪条，先读 §5 的内容信任边界。
1. 🛡️ 崩溃保护（强烈推荐）
# 用 fatal-guard 包装你的进程，崩溃时自动捕获墓碑
npx @misaka-net/fatal-guard -- <your-command>
# 墓碑自动转为 draft lesson
python3 scripts/tombstone_to_draft.py --from-file tombstone.json
# 或通过管道直连
fatal-guard -- node app.js 2>&1 | python3 scripts/tombstone_to_draft.py --stdin
崩溃 → 墓碑 JSON → draft lesson → bench-core 动态题库。你的每一次崩溃都在喂养整个网络。
2. 核心工作流：遇到问题先检索
新用户无需 clone：加
--remote直接查 D1 服务，免注册、不限次数（只有反爬突发保护）。
python3 search_knowledge.py "你的关键词" --remote # 远程直查（推荐，无需 clone）
python3 search_knowledge.py "关键词" # 本地 BM25（需 clone）
python3 search_knowledge.py "关键词" --lessons # 只看 lessons
python3 search_knowledge.py "关键词" --ref # 只看 reference
python3 search_knowledge.py "关键词" --titles # 只看标题
Agent 侧更常用 MCP：misakanet_search → misakanet_get_lesson → （无命中时）
misakanet_submit_intake。
3. 🔌 MCP 接口：端点 / 工具 / 注册 / streaming
端点：https://misakanet.org/mcp（MCP Streamable HTTP 传输，JSON-RPC 2.0）
3.1 三种调用形态
| 形态 | 请求 | 说明 |
|---|---|---|
| 普通 JSON | POST + Accept: application/json |
最常用；一次请求一个响应 |
| Streaming（SSE） | POST + Accept: application/json, text/event-stream |
服务端以 event: message 分块返回；长任务/逐块消费用，curl 加 -N |
| SSE 长连接 | GET + Accept: text/event-stream |
保持打开的流（健康检查/持续事件）；方法用错会返回 405 并提示正确用法 |
两个要带的请求头（第二个是「带错值才失败」——实测 2026-09-12：缺席放行，非法值 403）：
-H 'MCP-Protocol-Version: 2025-06-18' # 协议版本
-H 'Origin: https://misakanet.org' # MCP 规范要求：防 DNS rebinding；实测缺席=200、非法值（如 https://evil.example.com）→ 403 invalid Origin
3.2 工具清单（8 个）
| 工具 | 用途 | 鉴权 |
|---|---|---|
misakanet_search |
按错误文本/关键词检索课程；detail 三档（compact 默认 / summary / full）；FAQ 命中也会返回；无命中时返回 no_match + 可直接调用的 intake 指引 |
开放（匿名不限次数；同一地址有突发上限） |
misakanet_get_lesson |
按 id 或 path 取单篇课程正文（单次 ≤5000 字符；超长会返回 truncated: true + content_length + full_content_url，自己判断要不要取全文——2026-09-24 前是静默截断，457 篇里有 53 篇被腰斩） |
开放（同上，共用一个突发窗口） |
misakanet_submit_intake |
匿名报料/提问（kind="missing_lesson" / "question" / "proposal"；省略时自动判定问题与故障，计划类必须显式传 proposal）→ 服务端去重后开 GitHub issue；原样重发同一段 problem 文本即可回拉：问句返回维护者答复（answered），报料被某篇课程引用后返回转换回执（converted / receipt / events，判据是课程 frontmatter 的 source: 与 provenance.issue，SSOT 是 scripts/intake_receipt.py）；每次响应都带 dedup_key 与 poll_hint（recheck_after_seconds：问题 6h、报料 24h） |
开放（限流，无需账号） |
misakanet_me_inbox |
查 intake 状态（answered/pending/converted/not_found），不重新提交。传 intake_id（仅 question）或 dedup_key（所有 intake） |
开放（限流，无需账号） |
misakanet_write_lesson |
结构化提交完整课程（title/domain/problem/root_cause/fix）→ 走 lesson-gate |
需 Authorization: Bearer mcp_... |
misakanet_preflight |
高风险操作前的风险检查 | 开放（只读，限流） |
misakanet_register |
注册匿名节点 → 返回 node_id + token |
开放 |
misakanet_me_events |
取"课程被复用"的证据（E4 = 观测到的复用：search_signals 里 ≥2 个不同 client_hint 真的检索到过这篇；基准引用、helpful 票、课程正文里的 verified by: 行都只是 E3 提示 —— 引用只是沙箱级证据，票只是提示，正文自述是课程在说自己，三者任何数量都不给 E4：票自 2026-10-16、正文自述自 2026-10-17） |
开放（刻意开放：复用前应能自由核验信任证据） |
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 Changed · +4 lines · +174 tokens per session 0acae4c8fb17
- 4d ago Changed · +3 lines · +72 tokens per session 77035159abf9
- 9d ago Changed · +28 tokens per session 6abb24706d2b
- 20d ago Changed · +2 lines · +59 tokens per session 9c424dac80f9
- 21d ago Changed · +15 lines · +425 tokens per session 24a8a7088345
- 22d ago Changed · +1 lines · +59 tokens per session 1199f7b1f577
- 1mo ago Changed · +9 tokens per session a91c86a0c2e6
- 1mo ago Changed · +18 tokens per session 065dda59b0d5
- 1mo ago Changed · +120 tokens per session 14876812f86d
- 1mo ago First seen · 213 lines · 4,414 tokens per session scan C c0333cab50b2
MisakaNet AGENTS.md is an instructions file published in the GitHub repository Ikalus1988/MisakaNet (554 stars, last pushed today), licensed Apache-2.0. It adds 5,378 tokens to every session, about $0.0215 per session on Opus 5.5. 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.
MisakaNet AGENTS.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| MisakaNet AGENTS.mdIkalus1988/MisakaNet | Instructions file | 554 | today | Apache-2.0 | C | 5,378 | Codex, OpenCode |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | today | MIT | A | 7,461 | Codex, OpenCode |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | yesterday | MIT | A | 4,680 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 5,516 | GitHub Copilot |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 274k | 2d ago | MIT | A | 4,313 | Codex, OpenCode |