dsh-model-everything 0.3.0 → 0.4.0

package/README.en-US.md

@@ -44,6 +44,21 @@ reduce the model first, and if it still explodes the property needs a tool that
 for partial-order reduction, NuSMV for symbolic search, CBMC for bounded SAT. This plugin builds no
 abstraction and reads no program source.
 
+## Report cards
+
+A report still prints in full, and gains a card where the UI can render one. The card is not a second
+conclusion: it is the same computation, arranged for a reader.
+
+- **The verdict first**, then the first finding, then the next step;
+- **A diff keeps two questions apart**: what this change added, and what the current report decides on its
+  own - they differ exactly when a clean change lands on an already-failing report;
+- **A design review keeps three levels apart**: a violation decides the verdict, a signal is a structural
+  fact, a fact is a measurement;
+- **A concurrency scan** says it routed the claims and **proved none of them**;
+- Without a card the raw text is what shows, so an unsupported UI loses nothing.
+
+A card **saves no tokens**: the model still gets the full text.
+
 ## What it covers
 
 | Domain | What it does |
@@ -215,12 +230,17 @@ The plugin ships as a cordis bundle, declared by `dsh.bundle` in the root `packa
 ```bash
 # from a source checkout (development)
 dsh plugin --profile <profile> add "file:$PWD"
-# the published version, from npm
-dsh plugin --profile <profile> add dsh-model-everything
+# the published version, from npm: pin it
+dsh plugin --profile <profile> add dsh-model-everything@<version>
 # without a global dsh
-npx -p @deepseek-ai/dsh dsh plugin --profile <profile> add dsh-model-everything
+npx -p @deepseek-ai/dsh dsh plugin --profile <profile> add dsh-model-everything@<version>
 ```
 
+**Why pin it.** Right after a release, an unpinned install resolves the previous version: the mirror's
+`latest` lags, and pnpm's `minimumReleaseAge` refuses a version that was published minutes ago. With the
+version named, the dsh installer records the exception in the profile's `minimumReleaseAgeExclude` itself;
+check the result with `--dump-config` and the installed copy's `package.json`.
+
 Restart the profile afterwards; `dsh --profile <profile> --dump-config` should list this plugin. Remove it with the same plugin manager.
 
 It registers its skills through the standard DSH filesystem skill provider and its tools on `ctx.tools`, plus one model-visible status entry that reports registration state and the model schema versions the bundled engines accept.

package/README.md

@@ -35,6 +35,18 @@ DeepSeek Harness 原生插件。把设计与代码里**关于行为的断言**
 
 **它判定的是小模型，不是程序。** 引擎显式枚举状态，模型炸开时会报 `truncated` 并点名未决的检查。那不是重试一个更大的 `maxStates` 就能解决的：先减模型（去掉性质不读的变量、合并对称状态），仍然炸开就换不靠枚举的工具——SPIN 的偏序归约、NuSMV 的符号化、CBMC 的有界 SAT。本插件不建立抽象，也不读程序源码。
 
+## 报告卡片
+
+每份报告在终端里照常显示全文，同时在支持渲染的界面里给一张卡片。这不是第二份结论：卡片是**同一次计算**的另一种排布。
+
+- **先给判定**，再给首条问题，最后给下一步；
+- **diff 把两件事分开写**：这次改动新增了什么，以及当前报告本身是什么判定——它们只会在「干净改动落到一份本来就失败的报告上」时不一致；
+- **设计审阅分三级**：violation 决定判定，signal 是结构事实，fact 是测量；三者不会合并成一个数字；
+- **交错扫描**读完只说明「已把断言路由给能判定它的工具」，卡片上写着它**没有证明任何一条**；
+- 没有卡片时退回原始文本，不会因为界面不支持而丢掉结论。
+
+卡片**不节省任何 token**：模型看到的仍是全文。
+
 ## 能力范围
 
 | 域 | 做什么 |
@@ -206,12 +218,14 @@ model_verify model=examples/control-dispatcher.json
 ```bash
 # 从源码目录安装（开发中）
 dsh plugin --profile <profile> add "file:$PWD"
-# 从 npm 安装（发布版）
-dsh plugin --profile <profile> add dsh-model-everything
+# 从 npm 安装（发布版）：**钉住版本**
+dsh plugin --profile <profile> add dsh-model-everything@<version>
 # 未全局安装 dsh 时
-npx -p @deepseek-ai/dsh dsh plugin --profile <profile> add dsh-model-everything
+npx -p @deepseek-ai/dsh dsh plugin --profile <profile> add dsh-model-everything@<version>
 ```
 
+**为什么要钉版本**：刚发布的版本，不钉版本会装到上一个。两个原因叠加——镜像源的 `latest` 同步有延迟，而 pnpm 的 `minimumReleaseAge` 会拒收刚发布的版本。钉住版本时 dsh 的安装器会自动把这一条写进 profile 的 `minimumReleaseAgeExclude`，装完用 `--dump-config` 与安装副本的 `package.json` 核对版本。
+
 装完重启该 profile，`dsh --profile <profile> --dump-config` 里应能看到本插件的行。卸载同样走插件管理器，从目标 profile 移除。
 
 技能走 DSH 标准的 filesystem skill provider 注册，工具挂在 `ctx.tools` 上。另有一条模型可见的状态目录条目，可查注册状态与引擎接受的 schema 版本。

package/lib/automaton-tool.js

@@ -9,7 +9,7 @@ import { defineTool } from '@deepseek-ai/dsh-tools';
 import { verifyAutomaton } from './automaton.js';
 import { optionRefusal } from './option-shape.js';
 import { renderReportText } from './report-view.js';
-import { reportCard, reportResultView } from './presentation.js';
+import { reportCallView, reportCard, reportResultView } from './presentation.js';
 export const MODEL_AUTOMATON_TOOL_NAME = 'model_automaton_verify';
 export const modelAutomatonTool = defineTool({
     name: MODEL_AUTOMATON_TOOL_NAME,
@@ -48,10 +48,11 @@ export const modelAutomatonTool = defineTool({
         },
         // The card a UI shows, separated from the text the model reads: a verdict and the first thing wrong,
         // never the whole report.
-        presentationMeta(_args, value) {
+        presentationMeta(args, value) {
             return reportCard(value, MODEL_AUTOMATON_TOOL_NAME);
         },
     },
+    presentCall: reportCallView(MODEL_AUTOMATON_TOOL_NAME),
     presentResult: reportResultView(MODEL_AUTOMATON_TOOL_NAME),
     timeoutMs: 10_000,
     isConcurrencySafe: () => true,

package/lib/automaton.js

@@ -234,9 +234,6 @@ export function verifyAutomaton(input, options = {}) {
     for (const [index, label] of dfa.labels.entries()) {
         if (label.length === 0)
             continue;
-        if (label.length === 1 && !reach[index]) {
-            findings.push({ code: 'AUT001_UNREACHABLE_STATE', severity: 'warning', message: `state "${label[0]}" cannot be reached from init` });
-        }
         if (!dfa.accept[index] && reach[index] && !canAccept[index]) {
             findings.push({ code: 'AUT002_TRAP_STATE', severity: 'warning', message: `state "${label[0]}" is reachable but can never reach an accepting state` });
         }
@@ -273,10 +270,13 @@ export function verifyAutomaton(input, options = {}) {
                 findings.push({ code: 'AUT009_UNDECIDED', severity: 'warning', message: 'the product search stopped at its state cap before deciding the relation, so equality is neither proven nor refuted', detail: 'the product of two automata can be exponential; reduce one of them, or decide the relation with a tool that searches symbolically' });
             }
             else if (relation === 'equivalent') {
-                if (missing !== undefined || extra !== undefined) {
-                    const word = missing ?? extra;
-                    witness = { word: word ?? [], reason: missing !== undefined ? 'accepted by the first automaton, rejected by the second' : 'accepted by the second automaton, rejected by the first' };
-                    findings.push({ code: 'AUT006_NOT_EQUIVALENT', severity: 'error', message: 'the two automata accept different languages', detail: `shortest witness: ${word === undefined || word.length === 0 ? 'the empty word' : word.join(' ')}; ${witness.reason}` });
+                if (missing !== undefined) {
+                    witness = { word: missing, reason: 'accepted by the first automaton, rejected by the second' };
+                    findings.push({ code: 'AUT006_NOT_EQUIVALENT', severity: 'error', message: 'the two automata accept different languages', detail: `shortest witness: ${missing.length === 0 ? 'the empty word' : missing.join(' ')}; ${witness.reason}` });
+                }
+                else if (extra !== undefined) {
+                    witness = { word: extra, reason: 'accepted by the second automaton, rejected by the first' };
+                    findings.push({ code: 'AUT006_NOT_EQUIVALENT', severity: 'error', message: 'the two automata accept different languages', detail: `shortest witness: ${extra.length === 0 ? 'the empty word' : extra.join(' ')}; ${witness.reason}` });
                 }
             }
             else if (relation === 'subset') {
@@ -291,11 +291,12 @@ export function verifyAutomaton(input, options = {}) {
             }
         }
     }
-    const errors = findings.filter((finding) => finding.severity === 'error').length;
+    const errorFindings = findings.filter((finding) => finding.severity === 'error');
+    const errors = errorFindings.length;
     const warnings = findings.filter((finding) => finding.severity === 'warning').length;
     const verdict = errors > 0 ? 'fail' : warnings > 0 ? 'pass_with_findings' : 'pass';
     const reason = errors > 0
-        ? `${String(errors)} error finding(s) (first: ${findings.find((finding) => finding.severity === 'error')?.code ?? 'unknown'})`
+        ? `${String(errors)} error finding(s) (first: ${errorFindings[0].code})`
         : warnings > 0
             ? `no error findings; ${String(warnings)} warning finding(s)`
             : 'every check passed';

package/lib/compose-tool.js

@@ -1,6 +1,6 @@
 import { defineTool } from '@deepseek-ai/dsh-tools';
 import { runCompositionVerification } from './engine.js';
-import { reportCard, reportResultView } from './presentation.js';
+import { reportCallView, reportCard, reportResultView } from './presentation.js';
 import { OptionShapeError, optionRefusal, requireArrayOption } from './option-shape.js';
 import { renderReportText } from './report-view.js';
 export const COMPOSE_TOOL_NAME = 'model_compose_verify';
@@ -56,7 +56,9 @@ export const modelComposeTool = defineTool({
         presentationMeta(_args, value) {
             return reportCard(value, COMPOSE_TOOL_NAME);
         },
-    }, presentResult: reportResultView(COMPOSE_TOOL_NAME),
+    },
+    presentCall: reportCallView(COMPOSE_TOOL_NAME),
+    presentResult: reportResultView(COMPOSE_TOOL_NAME),
     timeoutMs: 10_000,
     isConcurrencySafe: () => true,
     async execute(args) {

package/lib/concurrency-tool.js

@@ -1,7 +1,7 @@
 import { defineTool } from '@deepseek-ai/dsh-tools';
 import { runConcurrencyScan } from './concurrency.js';
 import { renderReportText } from './report-view.js';
-import { reportCard, reportResultView } from './presentation.js';
+import { reportCallView, reportCard, scanResultView } from './presentation.js';
 export const INTERLEAVING_SCAN_TOOL_NAME = 'model_interleaving_scan';
 /**
  * Model-visible DSH tool that mines design documents/plans for concurrency-related
@@ -36,7 +36,9 @@ export const modelConcurrencyScanTool = defineTool({
         presentationMeta(_args, value) {
             return reportCard(value, INTERLEAVING_SCAN_TOOL_NAME);
         },
-    }, presentResult: reportResultView(INTERLEAVING_SCAN_TOOL_NAME),
+    },
+    presentCall: reportCallView(INTERLEAVING_SCAN_TOOL_NAME),
+    presentResult: scanResultView,
     timeoutMs: 10_000,
     isConcurrencySafe: () => true,
     async execute(args) {

package/lib/data-tool.js

@@ -3,7 +3,7 @@ import { runDataVerification, DATA_ENGINE_SCHEMA_VERSION } from './data-engine.j
 import { oneOf } from './model-file.js';
 import { optionRefusal, requireArrayOption, requireObjectOption } from './option-shape.js';
 import { renderReportText } from './report-view.js';
-import { reportCard, reportResultView } from './presentation.js';
+import { reportCallView, reportCard, reportResultView } from './presentation.js';
 export const DATAMODEL_VERIFY_TOOL_NAME = 'model_datamodel_verify';
 /**
  * Model-visible DSH tool wrapping the bundled data-model verification engine.
@@ -65,7 +65,9 @@ export const modelDataModelVerifyTool = defineTool({
         presentationMeta(_args, value) {
             return reportCard(value, DATAMODEL_VERIFY_TOOL_NAME);
         },
-    }, presentResult: reportResultView(DATAMODEL_VERIFY_TOOL_NAME),
+    },
+    presentCall: reportCallView(DATAMODEL_VERIFY_TOOL_NAME),
+    presentResult: reportResultView(DATAMODEL_VERIFY_TOOL_NAME),
     timeoutMs: 10_000,
     isConcurrencySafe: () => true,
     async execute(args) {

package/lib/design-tool.js

@@ -1,6 +1,6 @@
 import { defineTool } from '@deepseek-ai/dsh-tools';
 import { REPORT_SCHEMAS } from './engine.js';
-import { reportCard, reportResultView } from './presentation.js';
+import { designResultView, reportCallView, reportCard } from './presentation.js';
 import { oneOf } from './model-file.js';
 import { runDesignReview } from './design-engine.js';
 import { renderReportText } from './report-view.js';
@@ -48,7 +48,9 @@ export const designTool = defineTool({
         presentationMeta(_args, value) {
             return reportCard(value, DESIGN_TOOL_NAME);
         },
-    }, presentResult: reportResultView(DESIGN_TOOL_NAME),
+    },
+    presentCall: reportCallView(DESIGN_TOOL_NAME),
+    presentResult: designResultView,
     timeoutMs: 10_000,
     isConcurrencySafe: () => true,
     async execute(args) {

package/lib/edit-tool.js

@@ -13,7 +13,7 @@ import { defineTool } from '@deepseek-ai/dsh-tools';
 import { diffReports } from './baseline.js';
 import { applyEdits, readModelFile } from './editor.js';
 import { REPORT_SCHEMAS, runVerification } from './engine.js';
-import { reportCard, reportResultView } from './presentation.js';
+import { editResultView, reportCallView, reportCard } from './presentation.js';
 import { optionRefusal } from './option-shape.js';
 import { renderReportText } from './report-view.js';
 export const MODEL_EDIT_TOOL_NAME = 'model_edit';
@@ -60,7 +60,9 @@ export const modelEditTool = defineTool({
         presentationMeta(_args, value) {
             return reportCard(value, MODEL_EDIT_TOOL_NAME);
         },
-    }, presentResult: reportResultView(MODEL_EDIT_TOOL_NAME),
+    },
+    presentCall: reportCallView(MODEL_EDIT_TOOL_NAME),
+    presentResult: editResultView(MODEL_EDIT_TOOL_NAME),
     timeoutMs: 20_000,
     isConcurrencySafe: () => false,
     async execute(args) {

package/lib/export-tool.js

@@ -1,6 +1,6 @@
 import { defineTool } from '@deepseek-ai/dsh-tools';
 import { REPORT_SCHEMAS, DEFAULT_HASH_SPEC, modelHash, refusalVerdict, validateModel, verdictOf } from './engine.js';
-import { reportCard, reportResultView } from './presentation.js';
+import { cardWithSource, reportCallView, reportCard, reportResultView } from './presentation.js';
 import { ExportError, exportModel } from './exporters.js';
 import { renderReportText } from './report-view.js';
 /** The model hash of an export's input, when the input validates (the success path always does). */
@@ -47,9 +47,12 @@ export const modelExportTool = defineTool({
         },
         // The card a UI shows, separated from the text the model reads.
         presentationMeta(_args, value) {
-            return reportCard(value, EXPORT_TOOL_NAME);
+            const report = value;
+            return cardWithSource(value, EXPORT_TOOL_NAME, typeof report.primary === 'string' ? report.primary : undefined, typeof report.format === 'string' ? 'artifact: ' + report.format : 'artifact');
         },
-    }, presentResult: reportResultView(EXPORT_TOOL_NAME),
+    },
+    presentCall: reportCallView(EXPORT_TOOL_NAME),
+    presentResult: reportResultView(EXPORT_TOOL_NAME),
     timeoutMs: 10_000,
     isConcurrencySafe: () => true,
     async execute(args) {

package/lib/presentation.js

@@ -21,10 +21,37 @@ export function reportCard(value, family) {
     const findings = array(report.findings)
         .map((entry) => record(entry))
         .filter((entry) => entry !== undefined);
+    // Two shapes live under one payload: the families with a top-level `findings` array, and the verify
+    // family, whose findings sit inside `checks` and whose counts live in `summary`. Reading only the first
+    // shape made a failing verification card say "0 error(s)" while its own verdict said fail.
+    const violations = array(report.violations);
+    const signals = array(report.signals);
+    const facts = array(report.facts);
+    if (findings.length === 0 && violations.length > 0) {
+        // The design family keeps three levels apart on purpose: a violation decides the verdict, a signal is a
+        // structural fact, and a fact is a measurement. Only the violations are findings here, or the card would
+        // show a signal with an error mark.
+        for (const entry of violations) {
+            const finding = record(entry);
+            if (finding !== undefined)
+                findings.push({ ...finding, severity: 'error' });
+        }
+    }
+    else if (findings.length === 0 && violations.length === 0 && array(report.violations).length === 0 && !Array.isArray(report.violations)) {
+        for (const check of array(report.checks)) {
+            for (const entry of array(record(check)?.findings)) {
+                const finding = record(entry);
+                if (finding !== undefined)
+                    findings.push(finding);
+            }
+        }
+    }
     const severity = (entry) => String(entry.severity ?? 'info');
     const errors = findings.filter((entry) => severity(entry) === 'error');
     const warnings = findings.filter((entry) => severity(entry) === 'warning');
     const summary = record(report.summary);
+    const summaryErrors = typeof summary?.errors === 'number' ? summary.errors : undefined;
+    const summaryWarnings = typeof summary?.warnings === 'number' ? summary.warnings : undefined;
     const checksRun = typeof summary?.checksRun === 'number'
         ? summary.checksRun
         : array(report.checks).length > 0
@@ -32,20 +59,47 @@ export function reportCard(value, family) {
             : undefined;
     const first = errors[0] ?? warnings[0];
     const next = array(report.nextSteps).map((step) => String(step))[0];
+    const route = routeFor(family);
+    const deltaSummary = record(record(report.reportDelta)?.summary) ?? record(report.summary);
+    const delta = deltaSummary === undefined
+        ? undefined
+        : {
+            added: Number(deltaSummary.added ?? 0),
+            removed: Number(deltaSummary.removed ?? 0),
+            changed: Number(deltaSummary.changed ?? 0),
+            addedErrors: Number(deltaSummary.addedErrors ?? 0),
+        };
+    const changes = array(report.changes)
+        .map((entry) => record(entry))
+        .filter((entry) => entry !== undefined)
+        .map((entry) => ({ kind: String(entry.kind ?? '?'), what: String(entry.what ?? ''), ...(typeof entry.detail === 'string' && entry.detail !== '' ? { detail: entry.detail } : {}) }));
+    const firstAddedRaw = (array(report.added).map((entry) => record(entry)).find((entry) => entry !== undefined && String(entry.severity ?? '') === 'error')
+        ?? array(report.added).map((entry) => record(entry)).find((entry) => entry !== undefined));
     return {
         ...(typeof report.verdict === 'string' ? { verdict: report.verdict } : {}),
         ...(typeof report.verdictReason === 'string' ? { verdictReason: shorten(report.verdictReason) } : {}),
-        errors: errors.length,
-        warnings: warnings.length,
+        errors: summaryErrors ?? errors.length,
+        warnings: summaryWarnings ?? warnings.length,
         ...(checksRun === undefined ? {} : { checksRun }),
         first: first === undefined
             ? null
             : { code: String(first.code ?? 'UNKNOWN'), severity: severity(first), message: shorten(String(first.message ?? '')) },
         next: next === undefined ? null : shorten(next),
         family,
-        ...(routeFor(family) === undefined
+        ...(route === undefined ? {} : { route: { owns: route.owns, instead: route.notFor.tool } }),
+        ...(changes.length === 0 ? {} : { changes }),
+        ...(delta === undefined ? {} : { delta }),
+        ...(firstAddedRaw === undefined ? {} : { firstAdded: { code: String(firstAddedRaw.code ?? '?'), severity: String(firstAddedRaw.severity ?? 'info'), message: shorten(String(firstAddedRaw.message ?? '')) } }),
+        ...(typeof report.currentVerdict === 'string' ? { currentVerdict: report.currentVerdict } : {}),
+        ...(typeof summary?.nodes === 'number' && typeof summary?.edges === 'number'
+            ? { graph: { nodes: summary.nodes, edges: summary.edges, cycles: Number(summary.cycles ?? 0), isolated: Number(summary.isolated ?? 0), dangling: Number(summary.dangling ?? 0) } }
+            : {}),
+        ...(violations.length === 0 && signals.length === 0 && facts.length === 0
             ? {}
-            : { route: { owns: routeFor(family)?.owns ?? '', instead: routeFor(family)?.notFor.tool ?? '' } }),
+            : { levels: { violations: violations.length, signals: signals.length, facts: facts.length, skipped: array(report.checksSkipped).length } }),
+        ...(typeof summary?.absoluteClaims === 'number'
+            ? { claims: { absolute: summary.absoluteClaims, keywords: Number(summary.keywords ?? 0), families: claimFamilies(findings) } }
+            : {}),
     };
 }
 /** The card body: four lines, in the order a reader asks the questions. */
@@ -58,24 +112,198 @@ export function cardBlocks(card) {
     lines.push(head + '  —  ' + counts);
     if (card.verdictReason !== undefined)
         lines.push(card.verdictReason);
-    if (card.first !== null) {
-        lines.push((card.first.severity === 'error' ? '✗ ' : '⚠ ') + card.first.code + ': ' + card.first.message);
+    if (card.first != null) {
+        // A finding with no message prints its code and stops: a colon with nothing after it reads as a
+        // truncated message rather than as an absent one.
+        const mark = card.first.severity === 'error' ? '✗ ' : '⚠ ';
+        lines.push(mark + card.first.code + (card.first.message.trim() === '' ? '' : ': ' + card.first.message));
     }
-    if (card.next !== null)
+    if (card.next != null)
         lines.push('next: ' + card.next);
+    if (card.source !== undefined) {
+        lines.push('');
+        const shown = card.source.lines.length;
+        lines.push(card.source.label + (card.source.total > shown ? ' (first ' + String(shown) + ' of ' + String(card.source.total) + ' lines)' : ''));
+        lines.push(...card.source.lines);
+    }
     return [{ type: 'text', text: lines.join('\n') }];
 }
 /**
  * `presentResult` for a report tool: render the card from the persisted payload, and fall back to the
  * raw content by returning undefined when there is none.
  */
+/** The last path segment, for a title. No `node:path` here: this runs on replay too. */
+function lastSegment(path) {
+    const parts = path.split(/[\\/]/).filter((part) => part !== '' && part !== '.');
+    return parts[parts.length - 1] ?? path;
+}
+/**
+ * The pending card for a report tool: a title naming the artifact, and the file to follow.
+ *
+ * From the arguments only, because a UI calls this during streaming and again on replay; the file has not
+ * been opened at that point and must not be.
+ */
+/**
+ * The card payload with the artifact previewed.
+ *
+ * A translation and a rendering both produce text a reader wants to see the shape of, and neither should
+ * paste the whole file into a card: the preview is bounded, and it says how much it left out. The report
+ * keeps the artifact in full for whoever needs it.
+ */
+export function cardWithSource(value, family, text, label, limit = 14) {
+    const card = reportCard(value, family);
+    if (text === undefined || text.trim() === '')
+        return card;
+    const all = text.split(String.fromCharCode(10));
+    const lines = all.slice(0, limit);
+    return { ...card, source: { label, lines, total: all.length } };
+}
+export function reportCallView(family) {
+    return (args) => {
+        const a = record(args) ?? {};
+        const path = ['modelPath', 'repoPath', 'path']
+            .map((key) => a[key])
+            .find((value) => typeof value === 'string' && value !== '');
+        const kind = family === 'model_edit' ? 'edit' : family.startsWith('uml_') ? 'search' : 'other';
+        const title = path === undefined ? family : family + '  ' + lastSegment(path);
+        return { card: 'generic', title, kind, ...(path === undefined ? {} : { locations: [{ path }] }) };
+    };
+}
+/**
+ * The card for `model_edit`: what changed, and what the change did to the findings.
+ *
+ * Both lists are already in the report, so this is a rendering of the same value rather than a second
+ * computation - the rule every card here follows.
+ */
+export function editCardBlocks(card) {
+    const changes = card.changes ?? [];
+    const lines = [];
+    const head = card.verdict === undefined ? card.family : card.verdict.toUpperCase();
+    lines.push(head + (changes.length === 0 ? '  —  no change was applied' : '  —  ' + String(changes.length) + ' change(s)'));
+    for (const change of changes.slice(0, 6)) {
+        lines.push('· ' + change.kind + ' ' + change.what + (change.detail === undefined ? '' : ' (' + change.detail + ')'));
+    }
+    if (changes.length > 6)
+        lines.push('· … ' + String(changes.length - 6) + ' more in the report');
+    if (card.delta !== undefined) {
+        lines.push('findings: added ' + String(card.delta.added) + ' (' + String(card.delta.addedErrors) + ' error) · removed ' + String(card.delta.removed) + ' · changed ' + String(card.delta.changed));
+    }
+    if (card.next != null)
+        lines.push('next: ' + card.next);
+    return [{ type: 'text', text: lines.join('\n') }];
+}
+/**
+ * The card for `report_diff`: does this change add anything, read as counts and the first added error.
+ *
+ * A diff's `verdict` is about the delta, so the card leads with the delta and never with the current
+ * report's own verdict - the two differ exactly when a clean change lands on an already-failing report.
+ */
+export function diffCardBlocks(card) {
+    const delta = card.delta;
+    const lines = [];
+    const head = card.verdict === undefined ? card.family : card.verdict.toUpperCase();
+    lines.push(head + '  —  added ' + String(delta?.added ?? 0) + ' (' + String(delta?.addedErrors ?? 0) + ' error) · removed ' + String(delta?.removed ?? 0) + ' · changed ' + String(delta?.changed ?? 0));
+    if (card.currentVerdict !== undefined)
+        lines.push('the current report on its own: ' + card.currentVerdict);
+    if (card.firstAdded !== undefined)
+        lines.push('✗ ' + card.firstAdded.code + ': ' + card.firstAdded.message);
+    if ((delta?.added ?? 0) === 0)
+        lines.push('nothing was added: this change is accepted by the "no new findings" rule');
+    if (card.next != null)
+        lines.push('next: ' + card.next);
+    return [{ type: 'text', text: lines.join('\n') }];
+}
+/** `presentResult` for `model_edit`, falling back to the generic card when the payload has no changes. */
+export function editResultView(family) {
+    return (_args, result) => {
+        const card = record(result.meta);
+        if (card === undefined || card === null)
+            return undefined;
+        const changeCount = card.changes === undefined ? '' : '  ·  ' + String(card.changes.length) + ' change(s)';
+        return { card: 'generic', title: family + '  ' + (card.verdict ?? '') + changeCount, content: editCardBlocks(card) };
+    };
+}
 export function reportResultView(family) {
     return (_args, result) => {
         const card = record(result.meta);
         if (card === undefined || card === null)
             return undefined;
-        if (card.verdict === undefined && card.next === null && card.first === null)
+        if (card.verdict === undefined && card.next == null && card.first == null)
             return undefined;
         return { card: 'generic', title: family + '  ' + (card.verdict ?? ''), content: cardBlocks(card) };
     };
 }
+/** How many findings per family, for the scan card: the reader wants the shape, not the list. */
+function claimFamilies(findings) {
+    const counts = new Map();
+    for (const finding of findings) {
+        const family = typeof finding.family === 'string' ? finding.family : 'other';
+        counts.set(family, (counts.get(family) ?? 0) + 1);
+    }
+    return [...counts.entries()].sort((a, b) => b[1] - a[1]).map(([family, count]) => ({ family, count }));
+}
+/** The card for `uml_structure_verify`: the graph's shape first, then the first finding. */
+export function structureCardBlocks(card) {
+    const lines = [];
+    const head = card.verdict === undefined ? card.family : card.verdict.toUpperCase();
+    lines.push(head + '  —  ' + String(card.errors) + ' error(s) · ' + String(card.warnings) + ' warning(s)');
+    if (card.graph !== undefined) {
+        lines.push('graph: ' + String(card.graph.nodes) + ' nodes · ' + String(card.graph.edges) + ' edges · ' + String(card.graph.cycles) + ' cycle(s) · ' + String(card.graph.isolated) + ' isolated');
+    }
+    if (card.first != null)
+        lines.push((card.first.severity === 'error' ? '✗ ' : '⚠ ') + card.first.code + ': ' + card.first.message);
+    if (card.next != null)
+        lines.push('next: ' + card.next);
+    return [{ type: 'text', text: lines.join(String.fromCharCode(10)) }];
+}
+/** The card for `model_design_review`: violations, then signals, then facts - never one number. */
+export function designCardBlocks(card) {
+    const lines = [];
+    const head = card.verdict === undefined ? card.family : card.verdict.toUpperCase();
+    lines.push(head + '  —  ' + String(card.errors) + ' violation(s) · ' + String(card.warnings) + ' signal(s)');
+    if (card.levels !== undefined) {
+        lines.push('levels: ' + String(card.levels.violations) + ' violation · ' + String(card.levels.signals) + ' signal · ' + String(card.levels.facts) + ' fact · ' + String(card.levels.skipped) + ' check(s) skipped');
+    }
+    if (card.first != null)
+        lines.push('✗ ' + card.first.code + ': ' + card.first.message);
+    if (card.levels !== undefined && card.levels.skipped > 0)
+        lines.push('a skipped check examined nothing: silence there is not a pass');
+    if (card.next != null)
+        lines.push('next: ' + card.next);
+    return [{ type: 'text', text: lines.join(String.fromCharCode(10)) }];
+}
+/** The card for `model_interleaving_scan`: claims by family, and the reminder that nothing was proven. */
+export function scanCardBlocks(card) {
+    const lines = [];
+    const head = card.verdict === undefined ? card.family : card.verdict.toUpperCase();
+    lines.push(head + '  —  ' + String(card.errors) + ' absolute claim(s) · ' + String(card.warnings) + ' risk(s)');
+    if (card.claims !== undefined) {
+        const families = card.claims.families.map((entry) => entry.family + ' ' + String(entry.count)).join(' · ');
+        lines.push('claims: ' + String(card.claims.absolute) + ' absolute, ' + String(card.claims.keywords) + ' keyword(s) hit' + (families === '' ? '' : ' · ' + families));
+    }
+    if (card.first != null)
+        lines.push('✗ ' + card.first.code + ': ' + card.first.message);
+    lines.push('this plugin routed the claims; it proved none of them');
+    return [{ type: 'text', text: lines.join(String.fromCharCode(10)) }];
+}
+/** One `presentResult` per card body, since the shapes differ but the fallback does not. */
+export function familyResultView(family, body) {
+    return (_args, result) => {
+        const card = record(result.meta);
+        if (card === undefined || card === null)
+            return undefined;
+        return { card: 'generic', title: family + '  ' + (card.verdict ?? ''), content: body(card) };
+    };
+}
+export const structureResultView = familyResultView('uml_structure_verify', structureCardBlocks);
+export const designResultView = familyResultView('model_design_review', designCardBlocks);
+export const scanResultView = familyResultView('model_interleaving_scan', scanCardBlocks);
+/** `presentResult` for `report_diff`: the delta leads, never the current report's own verdict. */
+export function diffResultView(family) {
+    return (_args, result) => {
+        const card = record(result.meta);
+        if (card === undefined || card === null)
+            return undefined;
+        return { card: 'generic', title: family + '  ' + (card.verdict ?? ''), content: diffCardBlocks(card) };
+    };
+}

package/lib/report-diff-tool.js

@@ -1,7 +1,7 @@
 import { defineTool } from '@deepseek-ai/dsh-tools';
 import { diffReports } from './baseline.js';
 import { refusalVerdict } from './engine.js';
-import { reportCard, reportResultView } from './presentation.js';
+import { diffResultView, reportCallView, reportCard } from './presentation.js';
 import { renderReportText } from './report-view.js';
 export const REPORT_DIFF_TOOL_NAME = 'report_diff';
 /**
@@ -54,7 +54,9 @@ export const reportDiffTool = defineTool({
         presentationMeta(_args, value) {
             return reportCard(value, REPORT_DIFF_TOOL_NAME);
         },
-    }, presentResult: reportResultView(REPORT_DIFF_TOOL_NAME),
+    },
+    presentCall: reportCallView(REPORT_DIFF_TOOL_NAME),
+    presentResult: diffResultView(REPORT_DIFF_TOOL_NAME),
     timeoutMs: 10_000,
     isConcurrencySafe: () => true,
     async execute(args) {

package/lib/structure-tool.js

@@ -5,7 +5,7 @@ import { scanDependencyGraph } from './scan.js';
 import { reviewGranularity } from './granularity.js';
 import { OptionShapeError, optionRefusal, requireArrayOption } from './option-shape.js';
 import { renderReportText } from './report-view.js';
-import { reportCard, reportResultView } from './presentation.js';
+import { reportCallView, reportCard, structureResultView } from './presentation.js';
 export const STRUCTURE_TOOL_NAME = 'uml_structure_verify';
 /**
  * Model-visible DSH tool that audits a *structure* diagram — components, packages,
@@ -59,7 +59,8 @@ export const umlStructureTool = defineTool({
             return reportCard(value, STRUCTURE_TOOL_NAME);
         },
     },
-    presentResult: reportResultView(STRUCTURE_TOOL_NAME),
+    presentCall: reportCallView(STRUCTURE_TOOL_NAME),
+    presentResult: structureResultView,
     timeoutMs: 10_000,
     isConcurrencySafe: () => true,
     async execute(args) {

package/lib/tool.js

@@ -1,7 +1,7 @@
 import { defineTool } from '@deepseek-ai/dsh-tools';
 import { runVerification } from './engine.js';
 import { renderReportText } from './report-view.js';
-import { reportCard, reportResultView } from './presentation.js';
+import { reportCallView, reportCard, reportResultView } from './presentation.js';
 import { OptionShapeError, optionRefusal, requireObjectOption } from './option-shape.js';
 import { oneOf } from './model-file.js';
 export const MODEL_VERIFY_TOOL_NAME = 'model_verify';
@@ -61,10 +61,11 @@ export const modelVerifyTool = defineTool({
             return [{ type: 'text', text: renderReportText(value, { full: args.reportView === 'full', check: args.check }) }];
         },
         // The card a UI shows, separated from the text the model reads.
-        presentationMeta(_args, value) {
+        presentationMeta(args, value) {
             return reportCard(value, MODEL_VERIFY_TOOL_NAME);
         },
     },
+    presentCall: reportCallView(MODEL_VERIFY_TOOL_NAME),
     presentResult: reportResultView(MODEL_VERIFY_TOOL_NAME),
     timeoutMs: 10_000,
     isConcurrencySafe: () => true,