@pathmode/mcp-server 1.5.0 → 1.6.0

package/dist/index.js

@@ -33903,6 +33903,7 @@ exports.getCompileIntentPrompt = getCompileIntentPrompt;
 exports.formatIntentMd = formatIntentMd;
 exports.formatCursorRules = formatCursorRules;
 exports.formatClaudeMdSection = formatClaudeMdSection;
+exports.formatOutcomeRubric = formatOutcomeRubric;
 /** Extract text from a string or structured outcome. */
 function getOutcomeText(o) {
     return typeof o === 'string' ? o : o.text;
@@ -34258,6 +34259,224 @@ function formatClaudeMdSection(spec) {
     sections.push('<!-- PATHMODE:END -->');
     return sections.join('\n\n');
 }
+function getOutcomePriority(o) {
+    return typeof o === 'string' ? undefined : o.priority;
+}
+/** Map an outcome priority to grader-facing strength language. */
+function priorityToBar(priority) {
+    switch (priority) {
+        case 'should': return 'Expected';
+        case 'could': return 'Optional — note if present, do not fail if absent';
+        case 'must':
+        default: return 'Required';
+    }
+}
+/**
+ * Dedup constraint strings by a normalized key — strips trailing parenthetical
+ * refs (e.g. "(CON-2)"), trailing punctuation, and case, so an intent constraint
+ * and its constitution twin collapse to one. Keeps the first (more specific) form.
+ */
+function dedupeByNormalized(items) {
+    const seen = new Set();
+    const out = [];
+    for (const raw of items) {
+        const item = raw?.trim();
+        if (!item)
+            continue;
+        const key = item
+            .toLowerCase()
+            .replace(/\s*\([^)]*\)\s*/g, ' ')
+            .replace(/[.\s]+$/g, '')
+            .replace(/\s+/g, ' ')
+            .trim();
+        if (seen.has(key))
+            continue;
+        seen.add(key);
+        out.push(item);
+    }
+    return out;
+}
+/**
+ * Build the writer-facing task description — what the agent reads. Maps to the
+ * `description` field of a `user.define_outcome` event.
+ */
+function buildWriterDescription(spec) {
+    const sections = [];
+    sections.push(`Deliver: ${spec.title || 'Untitled intent'}.`);
+    if (spec.objective) {
+        sections.push('');
+        sections.push(`Why this matters: ${spec.objective}`);
+    }
+    if (spec.scope?.inScope?.length) {
+        sections.push('');
+        sections.push('In scope:');
+        for (const item of spec.scope.inScope)
+            sections.push(`- ${item}`);
+    }
+    if (spec.scope?.outOfScope?.length) {
+        sections.push('');
+        sections.push('Out of scope — do not do these:');
+        for (const item of spec.scope.outOfScope)
+            sections.push(`- ${item}`);
+    }
+    const ic = spec.implementationContext;
+    if (ic?.relevantAreas?.length) {
+        sections.push('');
+        sections.push('Relevant areas of the codebase:');
+        for (const a of ic.relevantAreas) {
+            if (a?.path?.trim())
+                sections.push(`- ${a.path}${a.reason?.trim() ? ` — ${a.reason}` : ''}`);
+        }
+    }
+    if (ic?.currentBehavior?.trim()) {
+        sections.push('');
+        sections.push(`Current behavior: ${ic.currentBehavior.trim()}`);
+    }
+    sections.push('');
+    sections.push('Write your deliverables to /mnt/session/outputs/. Iterate until the grader is satisfied.');
+    return sections.join('\n');
+}
+/**
+ * Build the grader-facing rubric — what the independent grader reads. Maps to the
+ * `rubric` field of a `user.define_outcome` event. Every criterion is written to
+ * force the grader to find concrete evidence rather than trust the writer.
+ */
+function buildGraderRubric(spec, opts = {}) {
+    const sections = [];
+    sections.push(`You are grading the artifact(s) in /mnt/session/outputs/ for: "${spec.title || 'Untitled intent'}".`);
+    sections.push('Score each criterion below independently. PASS a criterion only when you can point to concrete evidence in the artifact — a measured value, a visible state, a passing test, or a quoted line. The writer asserting it is done is NOT evidence.');
+    // Outcomes → coverage criteria
+    const outcomes = (spec.outcomes ?? []).filter((o) => getOutcomeText(o)?.trim());
+    if (outcomes.length) {
+        sections.push('');
+        sections.push('## Outcomes (coverage)');
+        let i = 1;
+        for (const o of outcomes) {
+            const bar = priorityToBar(getOutcomePriority(o));
+            sections.push(`${i}. [${bar}] ${getOutcomeText(o)}`);
+            sections.push('   - Evidence: point to the specific value, state, or test result in the artifact that proves this. If you cannot, mark FAIL.');
+            i++;
+        }
+    }
+    // Edge cases → must-handle criteria (structured edge cases + implementation-context risks)
+    const edgeCases = (spec.edgeCases ?? []).filter((ec) => ec.scenario?.trim() || ec.expectedBehavior?.trim());
+    const risks = (spec.implementationContext?.risks ?? []).filter((r) => r?.trim());
+    if (edgeCases.length || risks.length) {
+        sections.push('');
+        sections.push('## Edge cases (must handle)');
+        for (const ec of edgeCases) {
+            sections.push(`- ${ec.scenario} → expected: ${ec.expectedBehavior}`);
+            sections.push(`   - Evidence: locate or construct the "${ec.scenario}" condition and confirm the expected behavior occurs. Absent handling = FAIL.`);
+        }
+        for (const r of risks) {
+            sections.push(`- Guard against: ${r}`);
+            sections.push('   - Evidence: show the artifact handles this failure mode — a test, a guard clause, or a visible safe state. Unaddressed = FAIL.');
+        }
+    }
+    // Constraints + constitution + out-of-scope → out-of-bounds / no-fire list
+    const outOfBounds = [];
+    if (spec.constraints?.length)
+        outOfBounds.push(...spec.constraints);
+    if (opts.constitutionRules?.length)
+        outOfBounds.push(...opts.constitutionRules);
+    if (spec.scope?.outOfScope?.length) {
+        for (const item of spec.scope.outOfScope)
+            outOfBounds.push(`Stays out of scope: ${item}`);
+    }
+    const cleanedBounds = dedupeByNormalized(outOfBounds);
+    if (cleanedBounds.length) {
+        sections.push('');
+        sections.push('## Constraints (out of bounds — FAIL the artifact if any are violated)');
+        for (const c of cleanedBounds)
+            sections.push(`- ${c}`);
+    }
+    // Verification → procedures the grader must run to produce evidence
+    const v = spec.verification;
+    const checks = [];
+    for (const t of v?.e2eTests ?? [])
+        if (t?.trim())
+            checks.push(`[e2e] ${t}`);
+    for (const t of v?.unitTests ?? [])
+        if (t?.trim())
+            checks.push(`[unit] ${t}`);
+    for (const t of v?.manualChecks ?? [])
+        if (t?.trim())
+            checks.push(`[manual] ${t}`);
+    for (const t of spec.implementationContext?.verificationSuggestions ?? [])
+        if (t?.trim())
+            checks.push(`[suggested] ${t}`);
+    if (checks.length) {
+        sections.push('');
+        sections.push('## Checks to run (produce the evidence yourself)');
+        for (const c of checks)
+            sections.push(`- ${c}`);
+    }
+    // Health metrics → observable-signal criteria
+    const metrics = (spec.healthMetrics ?? []).filter((m) => m?.trim());
+    if (metrics.length) {
+        sections.push('');
+        sections.push('## Observable signals');
+        for (const m of metrics) {
+            sections.push(`- The artifact leaves a way to measure: ${m}`);
+        }
+    }
+    // Grader output format — proven shape from the Outcomes cookbook
+    sections.push('');
+    sections.push('## Output format');
+    sections.push('Line 1: a scoreboard — "Outcomes X/Y met. Constraints OK|VIOLATED. Edge cases X/Y."');
+    sections.push('Then one bullet per FAILED item only: "<section> <item> — FAIL. <what is missing and what to change>." One sentence per bullet.');
+    sections.push('Do not fail the artifact for style preferences, pre-existing issues outside this intent, or anything not listed above.');
+    return sections.join('\n');
+}
+/**
+ * Generate a Claude Managed Agents "Outcomes" rubric document for an intent.
+ *
+ * Outcomes (`user.define_outcome`) takes two separate fields: a `description`
+ * the writer agent reads, and a `rubric` the independent grader reads. This
+ * exporter renders both — clearly separated so they paste straight into the API —
+ * turning the intent's outcomes, edge cases, constraints, constitution rules and
+ * verification into checkable, evidence-forcing grader criteria.
+ *
+ * Docs: https://platform.claude.com/docs/en/managed-agents/define-outcomes
+ */
+function formatOutcomeRubric(spec, opts = {}) {
+    const maxIterations = opts.maxIterations ?? 5;
+    const description = buildWriterDescription(spec);
+    const rubric = buildGraderRubric(spec, opts);
+    const hasVerification = [
+        ...(spec.verification?.e2eTests ?? []),
+        ...(spec.verification?.unitTests ?? []),
+        ...(spec.verification?.manualChecks ?? []),
+    ].some((t) => t?.trim());
+    const doc = [];
+    doc.push('<!-- Pathmode → Claude Managed Agents: Outcomes rubric -->');
+    doc.push(`<!-- Generated ${new Date().toISOString()} | pathmode.io -->`);
+    doc.push('');
+    doc.push('# How to use this');
+    doc.push('');
+    doc.push('Send a `user.define_outcome` event to a Managed Agents session (beta header `managed-agents-2026-04-01`):');
+    doc.push('- Put **Writer Description** into the `description` field (the agent reads this).');
+    doc.push('- Put **Grader Rubric** into the `rubric` field (the independent grader reads this).');
+    doc.push(`- Suggested \`max_iterations\`: ${maxIterations} (Outcomes default 3, max 20).`);
+    if (!hasVerification) {
+        doc.push('');
+        doc.push('> ⚠ This intent has no verification defined, so grader criteria fall back to generic evidence requirements. Run `verify-intent` to sharpen the checks.');
+    }
+    doc.push('');
+    doc.push('---');
+    doc.push('');
+    doc.push('# Writer Description');
+    doc.push('');
+    doc.push(description);
+    doc.push('');
+    doc.push('---');
+    doc.push('');
+    doc.push('# Grader Rubric');
+    doc.push('');
+    doc.push(rubric);
+    doc.push('');
+    return doc.join('\n');
+}
 
 
 /***/ }),
@@ -64772,11 +64991,27 @@ function startMcpServer() {
             return { content: [{ type: 'text', text: `Graph analysis failed: ${e.message}` }] };
         }
     });
+    /** Map a cloud ApiIntent into the IntentFields shape the formatters consume. */
+    function apiIntentToFields(intent) {
+        const v = (intent.verification || {});
+        return {
+            id: intent.id,
+            title: intent.title,
+            objective: intent.objective,
+            outcomes: intent.outcomes ?? [],
+            constraints: intent.constraints,
+            edgeCases: (intent.edgeCases ?? []).map((e) => ({ scenario: e.scenario, expectedBehavior: e.expectedBehavior })),
+            healthMetrics: intent.healthMetrics,
+            scope: intent.scope,
+            verification: { manualChecks: v.manualChecks, unitTests: v.unitTests, e2eTests: v.e2eTests },
+            implementationContext: intent.implementationContext ?? undefined,
+        };
+    }
     server.registerTool('export_context', {
         title: 'Export Context',
-        description: 'Export workspace context as a formatted file. Use "claude-md" for CLAUDE.md (full workspace context), "cursorrules" for Cursor AI rules, or "intent-md" for a single intent specification file. For cursorrules/intent-md, product context is always derived from the resolved intent. For claude-md, pass productId to select a specific product, otherwise the first active product is used.',
+        description: 'Export workspace context as a formatted file. Use "claude-md" for CLAUDE.md (full workspace context), "cursorrules" for Cursor AI rules, "intent-md" for a single intent specification file, or "outcome-rubric" for a Claude Managed Agents Outcomes rubric (writer description + evidence-forcing grader rubric) derived from the resolved intent, its implementation context, and the workspace constitution. For cursorrules/intent-md/outcome-rubric, product context is always derived from the resolved intent. For claude-md, pass productId to select a specific product, otherwise the first active product is used.',
         inputSchema: {
-            format: zod_1.z.enum(['claude-md', 'cursorrules', 'intent-md']).describe('Export format'),
+            format: zod_1.z.enum(['claude-md', 'cursorrules', 'intent-md', 'outcome-rubric']).describe('Export format'),
             intentId: zod_1.z.string().optional().describe('Intent ID (optional, for cursorrules and intent-md)'),
             productId: zod_1.z.string().optional().describe('Product ID (optional, only used for claude-md format to select a specific product)'),
         },
@@ -64786,6 +65021,25 @@ function startMcpServer() {
             return { content: [{ type: 'text', text: 'Export requires cloud mode. Use PATHMODE_API_KEY to connect.' }] };
         }
         try {
+            if (format === 'outcome-rubric') {
+                const cloud = requireCloudClient();
+                const intent = intentId
+                    ? await cloud.getIntent(intentId)
+                    : (pickCurrentIntent(await cloud.listIntents('approved')) || pickCurrentIntent(await cloud.listIntents()));
+                if (!intent) {
+                    return { content: [{ type: 'text', text: 'No intent found to export. Pass an intentId, or create an approved intent first.' }] };
+                }
+                let constitutionRules = [];
+                try {
+                    const constitution = await cloud.getConstitution();
+                    constitutionRules = (constitution?.rules ?? [])
+                        .filter((r) => r?.isActive !== false && r?.text?.trim())
+                        .map((r) => r.text.trim());
+                }
+                catch { /* constitution is optional context */ }
+                const rubric = (0, intent_compiler_1.formatOutcomeRubric)(apiIntentToFields(intent), { constitutionRules });
+                return { content: [{ type: 'text', text: rubric }] };
+            }
             const content = await requireCloudClient().exportContext(format, intentId, productId);
             return { content: [{ type: 'text', text: content }] };
         }
@@ -65169,8 +65423,8 @@ function startMcpServer() {
                 }],
         };
     });
-    server.tool('intent_export', 'Export an intent spec as .cursorrules or CLAUDE.md section for AI agent consumption.', {
-        format: zod_1.z.enum(['cursorrules', 'claude-md']).describe('Export format'),
+    server.tool('intent_export', 'Export an intent spec as .cursorrules, a CLAUDE.md section, or a Claude Managed Agents Outcomes rubric for AI agent consumption.', {
+        format: zod_1.z.enum(['cursorrules', 'claude-md', 'outcome-rubric']).describe('Export format'),
         spec: zod_1.z.object(intentSpecSchema),
         path: zod_1.z.string().optional().describe('Output file path. Defaults to .cursorrules or CLAUDE.md'),
     }, async ({ format, spec, path }) => {
@@ -65185,6 +65439,17 @@ function startMcpServer() {
                     }],
             };
         }
+        else if (format === 'outcome-rubric') {
+            const content = (0, intent_compiler_1.formatOutcomeRubric)(spec);
+            const filePath = (0, path_1.resolve)(process.cwd(), path || 'outcome-rubric.md');
+            (0, fs_1.writeFileSync)(filePath, content, 'utf-8');
+            return {
+                content: [{
+                        type: 'text',
+                        text: `✓ Exported Outcomes rubric to ${filePath}\n\nPaste the Writer Description into the \`description\` field and the Grader Rubric into the \`rubric\` field of a Managed Agents \`user.define_outcome\` event. Docs: https://platform.claude.com/docs/en/managed-agents/define-outcomes`,
+                    }],
+            };
+        }
         else {
             const section = (0, intent_compiler_1.formatClaudeMdSection)(spec);
             const filePath = (0, path_1.resolve)(process.cwd(), path || 'CLAUDE.md');
@@ -65194,7 +65459,10 @@ function startMcpServer() {
                 existing = (0, fs_1.readFileSync)(filePath, 'utf-8');
             }
             catch { /* file doesn't exist yet */ }
-            const marker = /<!-- PATHMODE:START -->[\s\S]*?<!-- PATHMODE:END -->/;
+            // Tolerant of the suffixed start marker emitted by formatClaudeMdSection
+            // (`<!-- PATHMODE:START - Do not edit... -->`), so re-exports replace
+            // the block instead of appending a duplicate.
+            const marker = /<!-- PATHMODE:START[\s\S]*?-->[\s\S]*?<!-- PATHMODE:END -->/;
             const updated = marker.test(existing)
                 ? existing.replace(marker, section)
                 : existing ? existing + '\n\n' + section : section;

package/dist/intent-compiler.d.ts

@@ -32,6 +32,17 @@ export interface IntentFields {
         unitTests?: string[];
         e2eTests?: string[];
     };
+    /** Repo-analysis context (cloud intents). Richer than the structured fields:
+     *  risks are real edge cases, verificationSuggestions are sharp grader checks. */
+    implementationContext?: {
+        relevantAreas?: {
+            path: string;
+            reason: string;
+        }[];
+        currentBehavior?: string;
+        risks?: string[];
+        verificationSuggestions?: string[];
+    } | null;
 }
 /**
  * Returns the system prompt that turns Claude into a Socratic intent interviewer.
@@ -54,3 +65,25 @@ export declare function formatCursorRules(spec: IntentFields): string;
  * Adapted from lib/agentPromptGenerator.ts generateClaudeMdContent().
  */
 export declare function formatClaudeMdSection(spec: IntentFields): string;
+export interface OutcomeRubricOptions {
+    /**
+     * Workspace constitution rules. Rendered as cross-cutting "out of bounds"
+     * criteria the grader fails the artifact on if violated. Not present on the
+     * local IntentFields — supplied by the cloud export path.
+     */
+    constitutionRules?: string[];
+    /** Suggested grade-and-revise loop cap. Outcomes default is 3, max 20. */
+    maxIterations?: number;
+}
+/**
+ * Generate a Claude Managed Agents "Outcomes" rubric document for an intent.
+ *
+ * Outcomes (`user.define_outcome`) takes two separate fields: a `description`
+ * the writer agent reads, and a `rubric` the independent grader reads. This
+ * exporter renders both — clearly separated so they paste straight into the API —
+ * turning the intent's outcomes, edge cases, constraints, constitution rules and
+ * verification into checkable, evidence-forcing grader criteria.
+ *
+ * Docs: https://platform.claude.com/docs/en/managed-agents/define-outcomes
+ */
+export declare function formatOutcomeRubric(spec: IntentFields, opts?: OutcomeRubricOptions): string;

package/dist/intent-compiler.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"","sourceRoot":"","sources":["file:///Users/jannelammi/code/Pathmode/packages/mcp-server/src/intent-compiler.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAMH,MAAM,WAAW,YAAY;IACzB,EAAE,CAAC,EAAE,MAAM,CAAC;IACZ,KAAK,EAAE,MAAM,CAAC;IACd,SAAS,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,CAAC,MAAM,GAAG;QAAE,EAAE,CAAC,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAC;QAAC,QAAQ,CAAC,EAAE,MAAM,GAAG,QAAQ,GAAG,OAAO,CAAA;KAAE,CAAC,EAAE,CAAC;IAC7F,WAAW,CAAC,EAAE,MAAM,EAAE,CAAC;IACvB,SAAS,CAAC,EAAE;QAAE,QAAQ,EAAE,MAAM,CAAC;QAAC,gBAAgB,EAAE,MAAM,CAAA;KAAE,EAAE,CAAC;IAC7D,aAAa,CAAC,EAAE,MAAM,EAAE,CAAC;IACzB,KAAK,CAAC,EAAE;QAAE,OAAO,CAAC,EAAE,MAAM,EAAE,CAAC;QAAC,UAAU,CAAC,EAAE,MAAM,EAAE,CAAA;KAAE,CAAC;IACtD,YAAY,CAAC,EAAE;QACX,YAAY,CAAC,EAAE,MAAM,EAAE,CAAC;QACxB,SAAS,CAAC,EAAE,MAAM,EAAE,CAAC;QACrB,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;KACvB,CAAC;CACL;AAwDD;;;GAGG;AACH,wBAAgB,sBAAsB,IAAI,MAAM,CA2D/C;AAOD;;;GAGG;AACH,wBAAgB,cAAc,CAAC,IAAI,EAAE,YAAY,GAAG,MAAM,CA+FzD;AAOD;;;GAGG;AACH,wBAAgB,iBAAiB,CAAC,IAAI,EAAE,YAAY,GAAG,MAAM,CAwF5D;AAOD;;;;GAIG;AACH,wBAAgB,qBAAqB,CAAC,IAAI,EAAE,YAAY,GAAG,MAAM,CAsDhE"}
+{"version":3,"file":"","sourceRoot":"","sources":["file:///Users/jannelammi/code/Pathmode/packages/mcp-server/src/intent-compiler.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAMH,MAAM,WAAW,YAAY;IACzB,EAAE,CAAC,EAAE,MAAM,CAAC;IACZ,KAAK,EAAE,MAAM,CAAC;IACd,SAAS,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,CAAC,MAAM,GAAG;QAAE,EAAE,CAAC,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAC;QAAC,QAAQ,CAAC,EAAE,MAAM,GAAG,QAAQ,GAAG,OAAO,CAAA;KAAE,CAAC,EAAE,CAAC;IAC7F,WAAW,CAAC,EAAE,MAAM,EAAE,CAAC;IACvB,SAAS,CAAC,EAAE;QAAE,QAAQ,EAAE,MAAM,CAAC;QAAC,gBAAgB,EAAE,MAAM,CAAA;KAAE,EAAE,CAAC;IAC7D,aAAa,CAAC,EAAE,MAAM,EAAE,CAAC;IACzB,KAAK,CAAC,EAAE;QAAE,OAAO,CAAC,EAAE,MAAM,EAAE,CAAC;QAAC,UAAU,CAAC,EAAE,MAAM,EAAE,CAAA;KAAE,CAAC;IACtD,YAAY,CAAC,EAAE;QACX,YAAY,CAAC,EAAE,MAAM,EAAE,CAAC;QACxB,SAAS,CAAC,EAAE,MAAM,EAAE,CAAC;QACrB,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;KACvB,CAAC;IACF;sFACkF;IAClF,qBAAqB,CAAC,EAAE;QACpB,aAAa,CAAC,EAAE;YAAE,IAAI,EAAE,MAAM,CAAC;YAAC,MAAM,EAAE,MAAM,CAAA;SAAE,EAAE,CAAC;QACnD,eAAe,CAAC,EAAE,MAAM,CAAC;QACzB,KAAK,CAAC,EAAE,MAAM,EAAE,CAAC;QACjB,uBAAuB,CAAC,EAAE,MAAM,EAAE,CAAC;KACtC,GAAG,IAAI,CAAC;CACZ;AAwDD;;;GAGG;AACH,wBAAgB,sBAAsB,IAAI,MAAM,CA2D/C;AAOD;;;GAGG;AACH,wBAAgB,cAAc,CAAC,IAAI,EAAE,YAAY,GAAG,MAAM,CA+FzD;AAOD;;;GAGG;AACH,wBAAgB,iBAAiB,CAAC,IAAI,EAAE,YAAY,GAAG,MAAM,CAwF5D;AAOD;;;;GAIG;AACH,wBAAgB,qBAAqB,CAAC,IAAI,EAAE,YAAY,GAAG,MAAM,CAsDhE;AAOD,MAAM,WAAW,oBAAoB;IACjC;;;;OAIG;IACH,iBAAiB,CAAC,EAAE,MAAM,EAAE,CAAC;IAC7B,0EAA0E;IAC1E,aAAa,CAAC,EAAE,MAAM,CAAC;CAC1B;AA6KD;;;;;;;;;;GAUG;AACH,wBAAgB,mBAAmB,CAAC,IAAI,EAAE,YAAY,EAAE,IAAI,GAAE,oBAAyB,GAAG,MAAM,CAuC/F"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pathmode/mcp-server",
-  "version": "1.5.0",
+  "version": "1.6.0",
   "publishConfig": {
     "access": "public"
   },