misakanet 2.43.1 → 2.44.1

package/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "misakanet",
-  "version": "2.43.1",
+  "version": "2.44.1",
   "description": "Failure-memory network for AI coding agents: search evidence-rated failure lessons, submit new failures, and turn crashes into reusable knowledge.",
   "author": {
     "name": "Ikalus1988",

package/README.ja.md

@@ -12,7 +12,7 @@ mcp-name: io.github.Ikalus1988/misakanet
 > 誰かがすでに代償を支払ったバグを、セッションごとに一つずつ再発見せずに済みます —— 上部の **Lessons**
 > バッジが現在のコーパス規模です。
 >
-> エージェントネイティブなインターフェース: [MCP サーバー](https://misakanet.org/mcp)（7 ツール）、WebMCP（ブラウザの
+> エージェントネイティブなインターフェース: [MCP サーバー](https://misakanet.org/mcp)（8 ツール）、WebMCP（ブラウザの
 > `navigator.modelContext`）、`llms.txt` / `llms-full.txt`、そして `.well-known/agent-card.json` 経由の A2A ディスカバリー。
 
 <p align="center">

package/README.md

@@ -12,7 +12,7 @@ mcp-name: io.github.Ikalus1988/misakanet
 > the bugs someone already paid for, instead of rediscovering them one session at a time — the **Lessons**
 > badge above is the live corpus size.
 >
-> Agent-native interfaces: [MCP server](https://misakanet.org/mcp) (7 tools), WebMCP (browser
+> Agent-native interfaces: [MCP server](https://misakanet.org/mcp) (8 tools), WebMCP (browser
 > `navigator.modelContext`), `llms.txt` / `llms-full.txt`, and A2A discovery through
 > `.well-known/agent-card.json`.
 

package/README.zh-CN.md

@@ -439,7 +439,7 @@ npx @misaka-net/misakanet-setup
 
 | 层 | 是什么 | 缺了它会怎样 |
 |---|---|---|
-| **① 服务** | `https://misakanet.org/mcp`（Streamable HTTP，7 个工具，匿名**不限次数**，只有反爬突发保护）或**本地 stdio**（clone 后 `python3 scripts/mcp_server.py`，无限额）| 没有可查的地方 |
+| **① 服务** | `https://misakanet.org/mcp`（Streamable HTTP，8 个工具，匿名**不限次数**，只有反爬突发保护）或**本地 stdio**（clone 后 `python3 scripts/mcp_server.py`，无限额）| 没有可查的地方 |
 | **② 接入** | `npx @misaka-net/misakanet-setup`：把服务写进每个助手**自己的**配置文件（Claude Code / Codex / Hermes / OpenClaw / codewhale 各一套）| 你得自己知道 5 种配置文件分别怎么写 |
 | **③ 触发** | 规则块（「遇到报错先查经验库」）+ 检查点钩子（约 20 轮提醒沉淀）+ 14 天升级提示 | **端点在，但没有任何人会去调用它** |
 
@@ -452,8 +452,8 @@ setup 保证的是"工具确实在"和"该查的时刻更容易被抓住"，不
 
 ### 装完你得到什么（逐条可自检）
 
-1. **7 个 `misakanet_*` 工具出现在助手里** —— `codex mcp list` / `codewhale mcp tools` /
-   `claude mcp list` / `hermes mcp list`；**证据**：列表里有 `misakanet` 且 7 个工具；
+1. **8 个 `misakanet_*` 工具出现在助手里** —— `codex mcp list` / `codewhale mcp tools` /
+   `claude mcp list` / `hermes mcp list`；**证据**：列表里有 `misakanet` 且 8 个工具；
 2. **助手被要求「遇错先查」** —— 问一句「switch vision model」「context window exceeded」这类片段，它应该先说查过经验库；
    **证据**：事件流里出现 `misakanet_search`（claude/codewhale 用 `--output-format stream-json`，codex 用 `--json`）；
 3. **长会话会提醒沉淀** —— 约 20 轮后提醒把本次「失败 → 根因 → 修复 → 验证」变成一条课程

package/SKILL.md

@@ -45,8 +45,8 @@ Do NOT use MisakaNet for:
 > stdio server adds `misakanet_me_events` as a **proxy** of the hosted tool, so the reuse-evidence
 > steps below work on either install.
 >
-> * **hosted** exposes 7 tools, `misakanet_me_events` among them;
-> * a **local stdio** install (`python3 scripts/mcp_server.py`) exposes those same 7 **plus three that
+> * **hosted** exposes 8 tools, `misakanet_me_events` among them;
+> * a **local stdio** install (`python3 scripts/mcp_server.py`) exposes those same 8 **plus three that
 >   only make sense on your machine** — `misakanet_submit_usage`, `misakanet_usage_status` and
 >   `misakanet_memory_context`. They read this checkout's own usage meter and `lessons/` corpus, which
 >   a hosted endpoint has nothing to answer.

package/lib/client.js

@@ -178,6 +178,16 @@ window.__ModuleLoader__.load({
 			"panel.asked.voteNotHelpful": "not helpful",
 			"panel.asked.voteNone": "not voted",
 			"panel.asked.voteNoLesson": "no lesson",
+			"panel.asked.voteHint": "Same action as the verdict beside the assistant message, without scrolling back to it. Sends this lesson's id and one constant label (`dsh-web`); nothing else.",
+			"panel.asked.voteFailed": "could not reach misakanet.org — nothing was recorded",
+			"panel.asked.voteOne": "1 human confirmation — the reuse event now appears (E3: a reuse signal, not CI verification).",
+			"panel.asked.voteTwo": "2 human confirmations — enough for agents to read this as E4.",
+			"panel.asked.voteMore": "human confirmations — enough for agents to read this as E4.",
+			"panel.asked.noMatchTag": "no match",
+			"panel.asked.agentNext": "Hand this line to the agent:",
+			"panel.asked.agentNextCopy": "Copy the next step",
+			"panel.asked.agentNextText": "Please record this gap with the MisakaNet intake so the next person finds it: misakanet_submit_intake problem: ",
+			"panel.asked.agentNextWhy": "This page cannot file it: a report is re-checked by submitting the same text again, the server's dedup window is finite, and a page-triggered submit could open a second issue for one problem.",
 			"panel.counters": "this session · counters local to this browser · no account, nothing uploaded but your votes and reports",
 			"panel.report.title": "MisakaNet report",
 			"panel.report.coveredBy": "covered by ",
@@ -196,6 +206,9 @@ window.__ModuleLoader__.load({
 			"panel.trust.human": "human confirmation",
 			"panel.trust.noLevel": "no level",
 			"panel.trust.used": " · used ",
+			"panel.trust.external": "external reuse evidence: ",
+			"panel.trust.declared": "declared: ",
+			"panel.trust.e4signal": "live signal: E4",
 			"panel.trust.footnote": "The level is the lesson's own stored evidence. The count is other people's confirmations, not yours: one makes the reuse event appear, the second is what agents read as E4.",
 			"panel.activity.session": "this session: ",
 			"panel.activity.browser": "on this browser: ",
@@ -212,9 +225,11 @@ window.__ModuleLoader__.load({
 			"main.footnote": "The left column opens this page; the session-scoped counters live in the session's right-sidebar MisakaNet tab. The numbers here are what this browser has contributed.",
 			"tool.rawResult": "raw result",
 			"tool.readFailed": "result could not be read",
-			"tool.helpfulTitle": "Sends this lesson's id to misakanet.org as reuse evidence. Sends nothing else.",
+			"tool.helpfulTitle": "Sends this lesson's id and the fixed label `dsh-web` to misakanet.org as reuse evidence. Sends nothing else.",
 			"tool.notHelpfulTitle": "Sends this lesson's id and the search text, so the gap gets counted.",
-			"tool.helpfulHint": "Helpful sends the lesson id · not-helpful also sends the search text (gap stats)",
+			"tool.helpfulHint": "Helpful sends the lesson id and the label `dsh-web` · not-helpful also sends the search text (gap stats)",
+			"tool.verdict.helpful": "recorded — this lesson now has {count} human confirmation(s)",
+			"tool.verdict.helpfulSelf": "recorded — this lesson now has your human confirmation",
 			"tool.helpful": "👍 Helpful",
 			"tool.notHelpful": "👎 Not what I needed",
 			"tool.untitled": "(untitled)",
@@ -287,6 +302,16 @@ window.__ModuleLoader__.load({
 			"panel.asked.voteNotHelpful": "没用",
 			"panel.asked.voteNone": "未投票",
 			"panel.asked.voteNoLesson": "无课程",
+			"panel.asked.voteHint": "与助手消息旁的裁决行是同一个动作，不用再往回滚。只发这篇课程的 id 和一个固定标签（`dsh-web`），不发别的。",
+			"panel.asked.voteFailed": "连不上 misakanet.org —— 这次没有记下任何东西",
+			"panel.asked.voteOne": "1 个人工确认 —— 复用事件现在出现了（E3：复用信号，不是 CI 验证）。",
+			"panel.asked.voteTwo": "2 个人工确认 —— 已经是 agent 读到的 E4 了。",
+			"panel.asked.voteMore": "个人工确认 —— 已经是 agent 读到的 E4 了。",
+			"panel.asked.noMatchTag": "无命中",
+			"panel.asked.agentNext": "把这行交给 agent：",
+			"panel.asked.agentNextCopy": "复制下一步",
+			"panel.asked.agentNextText": "请用 MisakaNet 报料记下这个缺口，让下一个人能搜到：misakanet_submit_intake problem: ",
+			"panel.asked.agentNextWhy": "这个页面不能替你报：报料的复查方式是原样再提交一次，而服务端的去重窗口是有限的 —— 页面触发的提交可能为同一个问题开出第二个 issue。",
 			"panel.counters": "本会话 · 计数只存在本浏览器 · 没有账号，除了你的投票与报料，不上传任何东西",
 			"panel.report.title": "MisakaNet 报料",
 			"panel.report.coveredBy": "已被 ",
@@ -305,6 +330,9 @@ window.__ModuleLoader__.load({
 			"panel.trust.human": "人工确认",
 			"panel.trust.noLevel": "无等级",
 			"panel.trust.used": " · 已复用 ",
+			"panel.trust.external": "外部复用证据：",
+			"panel.trust.declared": "课程自述：",
+			"panel.trust.e4signal": "实时信号：E4",
 			"panel.trust.footnote": "等级来自课程自己存的证据；计数是别人的确认，不是你的：前者让复用事件出现，后者才是 agent 读到的 E4。",
 			"panel.activity.session": "本会话：",
 			"panel.activity.browser": "本浏览器：",
@@ -321,9 +349,11 @@ window.__ModuleLoader__.load({
 			"main.footnote": "左栏打开的就是这一页；会话级的计数在右栏的 MisakaNet 标签里。这里的数字是本浏览器贡献的部分。",
 			"tool.rawResult": "原始结果",
 			"tool.readFailed": "结果读不出来",
-			"tool.helpfulTitle": "把这篇课程的 id 发给 misakanet.org 作为复用证据，不发别的。",
+			"tool.helpfulTitle": "把这篇课程的 id 和一个固定标签（`dsh-web`）发给 misakanet.org 作为复用证据，不发别的。",
 			"tool.notHelpfulTitle": "把课程 id 和检索文本一起发出去，好把缺口统计进去。",
-			"tool.helpfulHint": "「有用」只发课程 id ·「没用」还会发检索文本（用于缺口统计）",
+			"tool.helpfulHint": "「有用」只发课程 id 与固定标签 `dsh-web` ·「没用」还会发检索文本（用于缺口统计）",
+			"tool.verdict.helpful": "已记录 —— 这篇课程现在有 {count} 个人工确认",
+			"tool.verdict.helpfulSelf": "已记录 —— 这篇课程现在有你的一次人工确认",
 			"tool.helpful": "👍 有用",
 			"tool.notHelpful": "👎 不是我需要的",
 			"tool.untitled": "（无标题）",
@@ -590,6 +620,32 @@ var DEFAULT_MCP_URL = "https://misakanet.org/mcp";
 		var OVERLAY_LIMIT = 5;
 		var OVERLAY_TIMEOUT_MS = 8000;
 
+		/**
+		 * The label every vote from this browser half carries: what cast it, not who.
+		 *
+		 * It is a **constant**, identical for every person and every session. There is no account, no
+		 * `client_id`, no cookie and no generated value anywhere in this file — the repository's privacy
+		 * line is "no account, no client_id, nothing uploaded but your votes and reports", and a fixed
+		 * surface label does not cross it while a generated id would. What it buys is the only thing that
+		 * was missing: `POST /api/helpful` used to accept `{lesson_id}` alone, so a vote cast from this
+		 * panel was byte-identical to one cast by `curl` and the panel's own effect was unmeasurable by
+		 * construction. The worker keeps `count` exactly as it was and writes this label to its own
+		 * per-source, per-day counter.
+		 */
+		var HELPFUL_SOURCE = "dsh-web";
+
+		/**
+		 * The exact line the panel hands back to the agent for a search that matched nothing.
+		 *
+		 * The panel cannot file it: a receipt is pulled only by submitting the same text again, and the
+		 * server's dedup window is finite, so a page-triggered submit could open a **second** issue for
+		 * one problem (acceptance item D4). A copyable string has no such power, and it is the one thing
+		 * this page can honestly offer — the agent in the chat is the only party that may submit.
+		 */
+		function intakeNextStep(query) {
+			return T("panel.asked.agentNextText") + JSON.stringify(String(query || ""));
+		}
+
 		function isJson(text) {
 			try { JSON.parse(text); return true; } catch (error) { return false; }
 		}
@@ -942,6 +998,23 @@ var DEFAULT_MCP_URL = "https://misakanet.org/mcp";
 			});
 		}
 
+		/**
+		 * The one place a helpful vote is posted, from either surface (the message verdict, the panel row).
+		 *
+		 * One function, so the two surfaces cannot drift into sending different bodies — and so the body
+		 * is stated once: this lesson's id, plus the constant `HELPFUL_SOURCE` label. No query text, no
+		 * identity, no timestamp of our own. The guard is a double-submit brake, not an auto-vote: it only
+		 * stops a *second* click on a sighting whose vote already landed, and abstaining remains free.
+		 */
+		function sendHelpfulVote(log, entry) {
+			if (!entry || entry.voted) return Promise.resolve(null);
+			return post("/api/helpful", { lesson_id: entry.lessonId, source: HELPFUL_SOURCE })
+				.then(function (answer) {
+					recordVote(log, entry, "helpful", answer);
+					return answer;
+				});
+		}
+
 		/**
 		 * The tool-call row for `misakanet_search`: what came back — and nothing else.
 		 *
@@ -974,7 +1047,11 @@ var DEFAULT_MCP_URL = "https://misakanet.org/mcp";
 			} else if (payload && payload.no_match) {
 				head.push(h("span", { key: "state" }, " " + T("panel.asked.noMatch")));
 			} else if (results.length) {
-				head.push(h("span", { key: "state" }, " " + results.length + (results.length === 1 ? " lesson" : " lessons")));
+				// Through the plural helper, like every other count: this line used to spell the plural by
+				// hand (`results.length === 1 ? " lesson" : " lessons"`), which is exactly the shape the
+				// helper exists to prevent — and it was also the one count in this file that could not be
+				// localized.
+				head.push(h("span", { key: "state" }, " " + count(results.length, T("unit.lesson"), T("unit.lessons"))));
 				head.push(h("span", { key: "top", style: DIM }, " · top: "));
 				head.push(h("span", { key: "title" }, top.title || top.id || T("tool.untitled")));
 				// Only fields the DEFAULT payload actually carries. `detail: "compact"` is
@@ -1031,21 +1108,23 @@ var DEFAULT_MCP_URL = "https://misakanet.org/mcp";
 			if (log.shownFor && log.shownFor !== messageId) return null;
 
 			function send(kind) {
-				// 👍 carries the lesson id and nothing else. 👎 has to carry the search text: that is what
-				// turns one annoyed person into a counted gap, and the line below says so before the click.
+				// 👍 carries the lesson id plus the fixed `dsh-web` label and nothing else — one call site
+				// (`sendHelpfulVote`), shared with the panel's row so the two surfaces cannot drift. 👎 has
+				// to carry the search text: that is what turns one annoyed person into a counted gap, and
+				// the line below says so before the click.
 				var call = kind === "helpful"
-					? post("/api/helpful", { lesson_id: pending.lessonId })
+					? sendHelpfulVote(log, pending)
 					: post("/api/feedback", { query: pending.query, lesson_id: pending.lessonId, feedback: "irrelevant" });
 				call.then(
 					function (answer) {
-						recordVote(log, pending, kind, answer);
+						if (kind !== "helpful") recordVote(log, pending, kind, answer);
 						setVerdict({
 							kind: kind,
 							count: answer && typeof answer.count === "number" ? answer.count : null,
 						});
 					},
 					function () {
-						setVerdict({ error: "could not reach misakanet.org — nothing was recorded" });
+						setVerdict({ error: T("panel.asked.voteFailed") });
 					}
 				);
 			}
@@ -1056,7 +1135,9 @@ var DEFAULT_MCP_URL = "https://misakanet.org/mcp";
 
 			if (verdict) {
 				var done = verdict.kind === "helpful"
-					? "recorded — this lesson now has " + (verdict.count === null ? "your" : verdict.count) + " human confirmation(s)"
+					? (verdict.count === null
+						? T("tool.verdict.helpfulSelf")
+						: T("tool.verdict.helpful", { count: verdict.count }))
 					: "recorded as not helpful — it feeds the unsolved-gap map";
 				return h("span", { style: Object.assign({}, DIM, { fontSize: "12px" }) }, done);
 			}
@@ -1151,6 +1232,11 @@ var DEFAULT_MCP_URL = "https://misakanet.org/mcp";
 			var trustState = react.useState({});
 			var trust = trustState[0];
 			var setTrust = trustState[1];
+			// What a click just did, per search row: the count the server answered with, or a failure. The
+			// point of change #2 is that the click is *consequential*, so the row has to say what moved.
+			var voteState = react.useState({});
+			var panelVotes = voteState[0];
+			var setPanelVotes = voteState[1];
 
 			react.useEffect(function () {
 				return subscribe(sessionId, function () {
@@ -1222,18 +1308,106 @@ var DEFAULT_MCP_URL = "https://misakanet.org/mcp";
 				count(intakes.length, "report") + " (" + converted + " became a lesson)"));
 
 			// ① what this session asked
+			//
+			// This is the session's own transcript, and each row already knows which searches were *not*
+			// voted — so this is where a vote costs one click instead of a scroll back to the finalized
+			// message the verdict lives beside. The row offers exactly the action the verdict row offers
+			// (`sendHelpfulVote`, one call site), and abstaining stays free: the button is an offer, it is
+			// not pre-selected and nothing posts without a click.
+			//
+			// A no-match row is drawn distinctly and carries the one thing this page may hand over: a
+			// copyable line for the agent. It cannot file the gap itself (see `intakeNextStep`).
+			function voteFromPanel(entry) {
+				var key = "q" + entry.callId + entry.at;
+				sendHelpfulVote(log, entry).then(
+					function (answer) {
+						var total = answer && typeof answer.count === "number" ? answer.count : null;
+						if (total !== null) {
+							// Written into the same per-lesson receipt section ③ renders, so the number the
+							// click produced and the number the section shows can never disagree.
+							asked.current[entry.lessonId] = true;
+							setTrust(function (prev) {
+								var next = Object.assign({}, prev);
+								next[entry.lessonId] = total;
+								return next;
+							});
+						}
+						var said = total === null ? ""
+							: (total === 1 ? T("panel.asked.voteOne")
+								: (total === 2 ? T("panel.asked.voteTwo")
+									: total + " " + T("panel.asked.voteMore")));
+						setPanelVotes(function (prev) {
+							var next = Object.assign({}, prev);
+							next[key] = said || T("panel.asked.voteFailed");
+							return next;
+						});
+					},
+					function () {
+						setPanelVotes(function (prev) {
+							var next = Object.assign({}, prev);
+							next[key] = T("panel.asked.voteFailed");
+							return next;
+						});
+					}
+				);
+			}
+
 			rows.push(h("div", { key: "s1", style: { marginTop: "12px", fontWeight: 600 } }, T("panel.asked")));
 			if (!searches.length) {
 				rows.push(h("div", { key: "s1e", style: DIM }, T("panel.asked.none")));
 			} else {
 				var searchRows = searches.slice().reverse().slice(0, 20).map(function (entry) {
-					var vote = entry.voted ? (entry.voteKind === "helpful" ? "Helpful" : "not helpful") : (entry.lessonId ? "not voted" : "no lesson");
-					return h("div", { key: "q" + entry.callId + entry.at, style: { borderTop: "1px solid " + MUTED, padding: "4px 0" } },
-						h("span", { style: DIM }, clockOf(entry.at) + "  "),
-						h("span", {}, entry.query || T("panel.asked.noQuery")),
-						h("span", { style: DIM }, "  " + (entry.noMatch ? T("panel.asked.nothingFound") : count(entry.hits, T("unit.lesson")))),
-						entry.title ? h("span", { style: DIM }, "  · top: " + entry.title + (entry.evidence ? " (" + entry.evidence + ")" : "")) : null,
-						h("span", { style: DIM }, "  · " + vote));
+					var key = "q" + entry.callId + entry.at;
+					var vote = entry.voted
+						? (entry.voteKind === "helpful" ? T("panel.asked.voteHelpful") : T("panel.asked.voteNotHelpful"))
+						: (entry.lessonId ? T("panel.asked.voteNone") : T("panel.asked.voteNoLesson"));
+					var said = panelVotes[key];
+					var missed = Boolean(entry.noMatch) || !entry.lessonId;
+					var controls = [];
+					if (entry.lessonId && !entry.voted) {
+						controls.push(h("button", {
+							key: "v", type: "button", style: BUTTON, title: T("panel.asked.voteHint"),
+							onClick: function () { voteFromPanel(entry); },
+						}, T("tool.helpful")));
+					}
+					if (missed) {
+						// A copy, never a submit. The string is the next step *for the agent*, because the
+						// only way to pull a receipt is to submit the same text again and a page-triggered
+						// re-check could open a second issue for one problem.
+						controls.push(h("button", {
+							key: "c", type: "button", style: BUTTON, title: intakeNextStep(entry.query),
+							onClick: function () {
+								copyText(intakeNextStep(entry.query), function (label) {
+									setPanelVotes(function (prev) {
+										var next = Object.assign({}, prev);
+										next[key] = label;
+										return next;
+									});
+								});
+							},
+						}, T("panel.asked.agentNextCopy")));
+					}
+					var lines = [
+						h("div", { key: "head", style: { borderLeft: "2px solid " + (missed ? "rgba(127,127,127,0.9)" : "transparent"), paddingLeft: missed ? "6px" : "8px" } },
+							h("span", { style: DIM }, clockOf(entry.at) + "  "),
+							h("span", {}, entry.query || T("panel.asked.noQuery")),
+							h("span", { style: DIM }, "  " + (entry.noMatch ? T("panel.asked.nothingFound") : count(entry.hits, T("unit.lesson")))),
+							missed ? h("span", { style: Object.assign({}, DIM, { border: "1px solid " + MUTED, borderRadius: "4px", padding: "0 4px", marginLeft: "4px", fontSize: "11px" }) },
+								T("panel.asked.noMatchTag")) : null,
+							entry.title ? h("span", { style: DIM }, "  · top: " + entry.title + (entry.evidence ? " (" + entry.evidence + ")" : "")) : null,
+							entry.lessonId ? h("span", { style: DIM }, "  · " + vote) : null),
+					];
+					if (missed) {
+						lines.push(h("div", { key: "next", style: DIM },
+							T("panel.asked.agentNext") + " ",
+							h("code", { style: { fontSize: "11px" } }, intakeNextStep(entry.query))));
+						lines.push(h("div", { key: "why", style: DIM }, T("panel.asked.agentNextWhy")));
+					}
+					if (controls.length || said) {
+						lines.push(h("div", { key: "do" }, controls,
+							said ? h("span", { style: DIM }, said) : null));
+					}
+					return h("div", { key: key, style: { borderTop: "1px solid " + MUTED, padding: "4px 0" } }, lines);
 				});
 				rows.push(h("div", { key: "s1r" }, searchRows));
 			}
@@ -1272,6 +1446,24 @@ var DEFAULT_MCP_URL = "https://misakanet.org/mcp";
 			}
 
 			// ③ how much these lessons are trusted
+			//
+			// Two different facts, kept apart because they come from two different places and mean two
+			// different things:
+			//   * **declared** — the lesson's own stored `evidence_level`, from the compact search payload;
+			//   * **external reuse evidence** — the confirmation count `GET /api/helpful?lesson_id=` returns,
+			//     i.e. other people's votes, not yours, which is the live signal `misakanet_me_events`
+			//     reports.
+			//
+			// What this panel canNOT show, stated here rather than implied: `misakanet_me_events` returns a
+			// richer receipt (per-event types, `count`/`queries`/`sources`, its own `evidence`) but it is an
+			// MCP tool, reachable only through the MCP endpoint with a JSON-RPC `tools/call` — and this
+			// browser half is forbidden from making one (acceptance D4: the same endpoint files issues). The
+			// count therefore carries the helpful-vote signal, and the regression citations and cross-node
+			// confirmations the tool also aggregates are **not** rendered. Nothing here is guessed: no field
+			// this endpoint does not return is displayed.
+			//
+			// The count itself is the same threshold `misakanet_me_events` reports: one helpful vote is a
+			// reuse signal (E3), two or more is E4 — which is also why the rule below is stated once.
 			rows.push(h("div", { key: "s3", style: { marginTop: "12px", fontWeight: 600 } }, T("panel.trust")));
 			if (!lessons.length) {
 				rows.push(h("div", { key: "s3e", style: DIM }, T("panel.trust.none")));
@@ -1285,10 +1477,15 @@ var DEFAULT_MCP_URL = "https://misakanet.org/mcp";
 					var counted = confirmations === undefined ? T("panel.trust.checking")
 						: (confirmations === null ? T("panel.trust.unavailable")
 							: count(confirmations, T("panel.trust.human")) + (confirmations >= 2 ? " → E4" : ""));
+					var declared = lesson.evidence || T("panel.trust.noLevel");
+					var external = h("span", { style: DIM },
+						T("panel.trust.external") + counted + (confirmations >= 2 ? " · " + T("panel.trust.e4signal") : ""));
 					return h("div", { key: "l" + lesson.lessonId, style: { borderTop: "1px solid " + MUTED, padding: "4px 0" } },
-						h("span", {}, lesson.title || lesson.lessonId),
-						h("span", { style: DIM }, "  " + (lesson.evidence || T("panel.trust.noLevel")) + " · " + counted +
-							T("panel.trust.used") + lesson.count + "× · " + timeRange(lesson)));
+						h("div", null,
+							h("span", {}, lesson.title || lesson.lessonId),
+							h("span", { style: DIM }, "  " + T("panel.trust.declared") + declared +
+								T("panel.trust.used") + lesson.count + "× · " + timeRange(lesson))),
+						h("div", null, external));
 				})));
 			}
 
@@ -2120,6 +2317,11 @@ var DEFAULT_MCP_URL = "https://misakanet.org/mcp";
 		}
 
 		exports.apply = apply;
+		// Test hooks, not surfaces: the headless render (`tests/test_dsh_plugin_panel_render.py`) has to
+		// seed the session log exactly the way the transcript rows do, or the panel it renders is an empty
+		// panel that proves nothing. Same constructor, same recorder, no second shape to drift.
+		exports.__sessionLog = sessionLog;
+		exports.__recordSearch = recordSearch;
 		exports.inject = inject;
 		exports.name = "misakanet-client";
 		return module.exports;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "misakanet",
-  "version": "2.43.1",
+  "version": "2.44.1",
   "license": "Apache-2.0",
   "description": "MisakaNet failure-memory skill — evidence-rated failure lessons for AI agents. DSH bundle: registers the MCP tools as mcp__misakanet__* (misakanet_search/get_lesson/…), served by the public Streamable HTTP endpoint https://misakanet.org/mcp — live tools from an npm install, in any profile, with no local python. A profile that prefers the repo's own stdio server can override the bundle row.",
   "icon": "icon.svg",

package/skills/misakanet/SKILL.md

@@ -45,8 +45,8 @@ Do NOT use MisakaNet for:
 > stdio server adds `misakanet_me_events` as a **proxy** of the hosted tool, so the reuse-evidence
 > steps below work on either install.
 >
-> * **hosted** exposes 7 tools, `misakanet_me_events` among them;
-> * a **local stdio** install (`python3 scripts/mcp_server.py`) exposes those same 7 **plus three that
+> * **hosted** exposes 8 tools, `misakanet_me_events` among them;
+> * a **local stdio** install (`python3 scripts/mcp_server.py`) exposes those same 8 **plus three that
 >   only make sense on your machine** — `misakanet_submit_usage`, `misakanet_usage_status` and
 >   `misakanet_memory_context`. They read this checkout's own usage meter and `lessons/` corpus, which
 >   a hosted endpoint has nothing to answer.