@shepai/cli 1.121.0-pr386.fb524e4 → 1.121.0-pr387.781b61f

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/evidence-output-parser.d.ts

@@ -4,6 +4,9 @@
  * Extracts structured Evidence records from free-form agent text output.
  * Looks for a fenced JSON code block containing an array of evidence objects.
  * Returns empty array gracefully on any parsing failure.
+ *
+ * Also provides validation to ensure UI-related evidence includes app-level
+ * proof (not just Storybook screenshots).
  */
 import type { Evidence } from '../../../../../domain/generated/output.js';
 /**
@@ -12,4 +15,22 @@ import type { Evidence } from '../../../../../domain/generated/output.js';
  * Returns empty array when no block found, JSON is malformed, or no valid records exist.
  */
 export declare function parseEvidenceRecords(output: string): Evidence[];
+export interface UiEvidenceValidationResult {
+    valid: boolean;
+    hasScreenshots: boolean;
+    hasAppScreenshots: boolean;
+    hasOnlyStorybookScreenshots: boolean;
+    warnings: string[];
+}
+/**
+ * Validate that UI-related evidence includes app-level proof, not just Storybook screenshots.
+ *
+ * Returns a validation result indicating whether the evidence set meets the requirement
+ * for app-level proof. Screenshots/videos are considered "Storybook-only" if their
+ * description or path contains storybook-related keywords and NO screenshots have
+ * app-level indicators.
+ *
+ * This validation is informational — it produces warnings but does not filter out records.
+ */
+export declare function validateUiEvidenceHasAppProof(evidence: Evidence[]): UiEvidenceValidationResult;
 //# sourceMappingURL=evidence-output-parser.d.ts.map

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/evidence-output-parser.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"evidence-output-parser.d.ts","sourceRoot":"","sources":["../../../../../../../../../packages/core/src/infrastructure/services/agents/feature-agent/nodes/evidence-output-parser.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAEH,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,2CAA2C,CAAC;AAwB1E;;;;GAIG;AACH,wBAAgB,oBAAoB,CAAC,MAAM,EAAE,MAAM,GAAG,QAAQ,EAAE,CAc/D"}
+{"version":3,"file":"evidence-output-parser.d.ts","sourceRoot":"","sources":["../../../../../../../../../packages/core/src/infrastructure/services/agents/feature-agent/nodes/evidence-output-parser.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAEH,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,2CAA2C,CAAC;AAwB1E;;;;GAIG;AACH,wBAAgB,oBAAoB,CAAC,MAAM,EAAE,MAAM,GAAG,QAAQ,EAAE,CAc/D;AAwCD,MAAM,WAAW,0BAA0B;IACzC,KAAK,EAAE,OAAO,CAAC;IACf,cAAc,EAAE,OAAO,CAAC;IACxB,iBAAiB,EAAE,OAAO,CAAC;IAC3B,2BAA2B,EAAE,OAAO,CAAC;IACrC,QAAQ,EAAE,MAAM,EAAE,CAAC;CACpB;AAED;;;;;;;;;GASG;AACH,wBAAgB,6BAA6B,CAAC,QAAQ,EAAE,QAAQ,EAAE,GAAG,0BAA0B,CAsC9F"}

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/evidence-output-parser.js

@@ -4,6 +4,9 @@
  * Extracts structured Evidence records from free-form agent text output.
  * Looks for a fenced JSON code block containing an array of evidence objects.
  * Returns empty array gracefully on any parsing failure.
+ *
+ * Also provides validation to ensure UI-related evidence includes app-level
+ * proof (not just Storybook screenshots).
  */
 import { EvidenceType } from '../../../../../domain/generated/output.js';
 // Matches a fenced JSON code block: ```json ... ```
@@ -46,3 +49,71 @@ export function parseEvidenceRecords(output) {
         return [];
     return parsed.filter(isValidRecord);
 }
+/**
+ * Patterns that indicate a screenshot is from Storybook rather than the actual app.
+ * Checked case-insensitively against description and relativePath.
+ */
+const STORYBOOK_PATTERNS = [/storybook/i, /story\b/i, /stories/i, /\b6006\b/];
+/**
+ * Patterns that indicate a screenshot is from the actual running application.
+ * Checked case-insensitively against description and relativePath.
+ */
+const APP_PROOF_PATTERNS = [/\bapp[:\-\s]/i, /dev\s*server/i, /running\s*app/i, /actual\s*app/i];
+function isScreenshot(e) {
+    return e.type === EvidenceType.Screenshot || e.type === EvidenceType.Video;
+}
+function matchesAny(text, patterns) {
+    return patterns.some((p) => p.test(text));
+}
+function isStorybookEvidence(e) {
+    return (matchesAny(e.description, STORYBOOK_PATTERNS) || matchesAny(e.relativePath, STORYBOOK_PATTERNS));
+}
+function isAppEvidence(e) {
+    // Explicitly marked as app-level evidence
+    if (matchesAny(e.description, APP_PROOF_PATTERNS) ||
+        matchesAny(e.relativePath, APP_PROOF_PATTERNS)) {
+        return true;
+    }
+    // A screenshot that is NOT storybook is treated as app-level evidence
+    return !isStorybookEvidence(e);
+}
+/**
+ * Validate that UI-related evidence includes app-level proof, not just Storybook screenshots.
+ *
+ * Returns a validation result indicating whether the evidence set meets the requirement
+ * for app-level proof. Screenshots/videos are considered "Storybook-only" if their
+ * description or path contains storybook-related keywords and NO screenshots have
+ * app-level indicators.
+ *
+ * This validation is informational — it produces warnings but does not filter out records.
+ */
+export function validateUiEvidenceHasAppProof(evidence) {
+    const screenshots = evidence.filter(isScreenshot);
+    const warnings = [];
+    if (screenshots.length === 0) {
+        return {
+            valid: true,
+            hasScreenshots: false,
+            hasAppScreenshots: false,
+            hasOnlyStorybookScreenshots: false,
+            warnings,
+        };
+    }
+    const appScreenshots = screenshots.filter(isAppEvidence);
+    const storybookScreenshots = screenshots.filter(isStorybookEvidence);
+    const hasOnlyStorybook = storybookScreenshots.length > 0 && appScreenshots.length === 0;
+    if (hasOnlyStorybook) {
+        warnings.push('UI evidence contains only Storybook screenshots. App-level screenshots from the running application are REQUIRED for UI features.');
+    }
+    if (screenshots.length > 0 && appScreenshots.length === 0 && storybookScreenshots.length === 0) {
+        // Screenshots exist but none clearly identified as app or storybook — warn to be explicit
+        warnings.push('UI screenshots lack clear app-level indicators. Include "app" in the description or filename to confirm they are from the actual running application.');
+    }
+    return {
+        valid: !hasOnlyStorybook,
+        hasScreenshots: screenshots.length > 0,
+        hasAppScreenshots: appScreenshots.length > 0,
+        hasOnlyStorybookScreenshots: hasOnlyStorybook,
+        warnings,
+    };
+}

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/evidence.node.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"evidence.node.d.ts","sourceRoot":"","sources":["../../../../../../../../../packages/core/src/infrastructure/services/agents/feature-agent/nodes/evidence.node.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;GAiBG;AAGH,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,+DAA+D,CAAC;AAEpG,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AAerD;;;;;GAKG;AACH,wBAAgB,kBAAkB,CAAC,QAAQ,EAAE,cAAc,IAG3C,OAAO,iBAAiB,KAAG,OAAO,CAAC,OAAO,CAAC,iBAAiB,CAAC,CAAC,CAuE7E"}
+{"version":3,"file":"evidence.node.d.ts","sourceRoot":"","sources":["../../../../../../../../../packages/core/src/infrastructure/services/agents/feature-agent/nodes/evidence.node.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;GAiBG;AAGH,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,+DAA+D,CAAC;AAEpG,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AAerD;;;;;GAKG;AACH,wBAAgB,kBAAkB,CAAC,QAAQ,EAAE,cAAc,IAG3C,OAAO,iBAAiB,KAAG,OAAO,CAAC,OAAO,CAAC,iBAAiB,CAAC,CAAC,CAkF7E"}

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/evidence.node.js

@@ -22,7 +22,7 @@ import { reportNodeStart } from '../heartbeat.js';
 import { recordPhaseStart, recordPhaseEnd } from '../phase-timing-context.js';
 import { updateNodeLifecycle } from '../lifecycle-context.js';
 import { buildEvidencePrompt } from './prompts/evidence-prompts.js';
-import { parseEvidenceRecords } from './evidence-output-parser.js';
+import { parseEvidenceRecords, validateUiEvidenceHasAppProof } from './evidence-output-parser.js';
 import { hasSettings, getSettings } from '../../../settings.service.js';
 /**
  * Factory that creates the evidence collection node function.
@@ -70,6 +70,15 @@ export function createEvidenceNode(executor) {
                 log.error(`Warning: evidence parsing failed: ${msg} — continuing with empty evidence`);
                 evidence = [];
             }
+            // --- Validate UI evidence includes app-level proof ---
+            const validationResult = validateUiEvidenceHasAppProof(evidence);
+            const validationMessages = [];
+            if (validationResult.warnings.length > 0) {
+                for (const warning of validationResult.warnings) {
+                    log.error(`Evidence validation: ${warning}`);
+                    validationMessages.push(`[evidence] Warning: ${warning}`);
+                }
+            }
             await recordPhaseEnd(timingId, durationMs);
             markPhaseComplete(state.specDir, 'evidence', log);
             return {
@@ -80,6 +89,7 @@ export function createEvidenceNode(executor) {
                     ...(evidence.length === 0
                         ? ['[evidence] Warning: no evidence records parsed from agent output']
                         : []),
+                    ...validationMessages,
                 ],
                 _needsReexecution: false,
             };

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/fast-implement.node.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"fast-implement.node.d.ts","sourceRoot":"","sources":["../../../../../../../../../packages/core/src/infrastructure/services/agents/feature-agent/nodes/fast-implement.node.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;GAUG;AAMH,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,+DAA+D,CAAC;AAEpG,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AAUrD;;;;;GAKG;AACH,wBAAgB,uBAAuB,CAAC,QAAQ,EAAE,cAAc,IAGhD,OAAO,iBAAiB,KAAG,OAAO,CAAC,OAAO,CAAC,iBAAiB,CAAC,CAAC,CA4D7E"}
+{"version":3,"file":"fast-implement.node.d.ts","sourceRoot":"","sources":["../../../../../../../../../packages/core/src/infrastructure/services/agents/feature-agent/nodes/fast-implement.node.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;GAUG;AAKH,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,+DAA+D,CAAC;AAEpG,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AAUrD;;;;;GAKG;AACH,wBAAgB,uBAAuB,CAAC,QAAQ,EAAE,cAAc,IAGhD,OAAO,iBAAiB,KAAG,OAAO,CAAC,OAAO,CAAC,iBAAiB,CAAC,CAAC,CAoD7E"}

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/fast-implement.node.js

@@ -9,7 +9,6 @@
  * Follows the same factory pattern as other nodes: takes executor
  * dependency, returns async (state) => Partial<FeatureAgentState>.
  */
-import { execSync } from 'node:child_process';
 import { mkdirSync, writeFileSync } from 'node:fs';
 import { join, dirname } from 'node:path';
 import { isGraphBubbleUp } from '@langchain/langgraph';
@@ -45,11 +44,6 @@ export function createFastImplementNode(executor) {
             const durationMs = Date.now() - startTime;
             const elapsed = (durationMs / 1000).toFixed(1);
             log.info(`Complete (${result.result.length} chars, ${elapsed}s)`);
-            // Validate that the executor actually produced file changes
-            const cwd = state.worktreePath || state.repositoryPath;
-            if (!hasWorktreeChanges(cwd)) {
-                throw new Error('[fast-implement] Agent produced no file changes — it may have entered plan mode or asked questions instead of implementing. Retrying.');
-            }
             // --- Evidence sub-agent: capture proof of completion (settings-gated) ---
             const evidenceEnabled = hasSettings() && getSettings().workflow.enableEvidence;
             const evidence = evidenceEnabled ? await collectEvidence(executor, state, log) : [];
@@ -113,21 +107,6 @@ async function collectEvidence(executor, state, log) {
         return [];
     }
 }
-/**
- * Check whether the worktree has any uncommitted changes (new, modified, or deleted files).
- * Uses `git status --porcelain` which outputs one line per changed file, or empty if clean.
- * Returns false if the git command fails (e.g. not a git repo).
- */
-function hasWorktreeChanges(cwd) {
-    try {
-        const output = execSync('git status --porcelain', { cwd, encoding: 'utf-8' });
-        return output.trim().length > 0;
-    }
-    catch {
-        // If git command fails, assume no changes (conservative — will trigger the error)
-        return false;
-    }
-}
 /**
  * Save evidence manifest to the shep home evidence folder so the
  * merge review UI can read it without accessing graph state.

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/prompts/evidence-prompts.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"evidence-prompts.d.ts","sourceRoot":"","sources":["../../../../../../../../../../packages/core/src/infrastructure/services/agents/feature-agent/nodes/prompts/evidence-prompts.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;GAWG;AAIH,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,gBAAgB,CAAC;AAExD,MAAM,WAAW,qBAAqB;IACpC,kEAAkE;IAClE,cAAc,EAAE,OAAO,CAAC;CACzB;AAED,wBAAgB,mBAAmB,CACjC,KAAK,EAAE,iBAAiB,EACxB,OAAO,GAAE,qBAAiD,GACzD,MAAM,CA8LR"}
+{"version":3,"file":"evidence-prompts.d.ts","sourceRoot":"","sources":["../../../../../../../../../../packages/core/src/infrastructure/services/agents/feature-agent/nodes/prompts/evidence-prompts.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;GAWG;AAIH,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,gBAAgB,CAAC;AAExD,MAAM,WAAW,qBAAqB;IACpC,kEAAkE;IAClE,cAAc,EAAE,OAAO,CAAC;CACzB;AAED,wBAAgB,mBAAmB,CACjC,KAAK,EAAE,iBAAiB,EACxB,OAAO,GAAE,qBAAiD,GACzD,MAAM,CA0MR"}

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/prompts/evidence-prompts.js

@@ -90,11 +90,17 @@ ${cwd}
 
 Review the completed tasks above and capture evidence for each.
 
-### MANDATORY Screenshot Rule
+### MANDATORY Screenshot Rule — App-Level Proof Required
 
-**If ANY task touches UI code (components, pages, views, styles, layouts, templates, JSX/TSX, HTML, CSS), you MUST capture at least one Screenshot.** This is non-negotiable. Changing a heading, button label, color, layout, or any visible element requires a screenshot proving the change renders correctly.
+**If ANY task touches UI code (components, pages, views, styles, layouts, templates, JSX/TSX, HTML, CSS), you MUST capture at least one Screenshot from the ACTUAL RUNNING APPLICATION (dev server).** This is non-negotiable. Changing a heading, button label, color, layout, or any visible element requires a screenshot proving the change renders correctly in the real app context.
 
-To capture screenshots:
+**CRITICAL: Storybook-only evidence is NOT sufficient for UI features.** Storybook screenshots may be included as supplementary evidence, but they do NOT replace app-level screenshots. You MUST capture screenshots from the actual running application (via the dev server) to prove the feature works in its real context with actual data, routing, and integration.
+
+For each UI change you MUST capture:
+1. **App screenshot (REQUIRED)**: A screenshot from the running dev server showing the feature in its actual app context. Include "app" or the actual page/route name in the screenshot file name (e.g., \`app-dashboard-new-toggle.png\`, \`app-settings-page.png\`).
+2. **Storybook screenshot (OPTIONAL, supplementary)**: If the component has a Storybook story, you may also capture a Storybook screenshot as supplementary evidence. Include "storybook" in the file name (e.g., \`storybook-toggle-component.png\`).
+
+To capture app-level screenshots:
 1. Install Playwright if not already available: \`npx playwright install chromium --with-deps 2>/dev/null || npx playwright install chromium\`
 2. Start the dev server in the background (e.g., \`npm run dev &\` or \`pnpm dev &\`)
 3. Wait for the server to be ready
@@ -114,11 +120,16 @@ To capture screenshots:
    \`\`\`
 5. Stop the dev server after capturing
 
+To capture Storybook screenshots (supplementary only):
+1. Start Storybook: \`npx storybook dev -p 6006 &\` or \`pnpm dev:storybook &\`
+2. Navigate to the relevant story and capture a screenshot
+3. Stop Storybook after capturing
+
 If Playwright installation fails, try alternative approaches (e.g., \`npx puppeteer\`, \`curl\` for API responses). Only skip screenshots if ALL browser automation approaches fail — and in that case, explicitly note the failure reason in the evidence description.
 
 ### Evidence by Task Type
 
-- **UI/Web tasks**: You MUST take screenshots (see mandatory rule above). Also run and capture test output.
+- **UI/Web tasks**: You MUST take screenshots FROM THE RUNNING APP (see mandatory rule above). Storybook-only screenshots are NOT acceptable as primary evidence. Also run and capture test output.
 - **Backend/API tasks**: Run the relevant test suite and capture test output showing passing tests.
 - **CLI tasks**: Run the CLI commands and capture terminal output demonstrating correct behavior.
 - **Refactoring/Infrastructure tasks**: Capture test output or build output confirming no regressions.
@@ -133,9 +144,10 @@ If Playwright installation fails, try alternative approaches (e.g., \`npx puppet
 ### Minimum Evidence Requirements
 
 Your output MUST include:
-1. At least one **Screenshot** if any task modifies UI/visual code (components, styles, templates, markup)
+1. At least one **Screenshot from the running app** (NOT Storybook) if any task modifies UI/visual code (components, styles, templates, markup). The screenshot description MUST clearly indicate it was captured from the actual application (e.g., "App: dashboard page showing new toggle", "App: settings page with updated layout"). Do NOT use descriptions that suggest Storybook-only proof.
 2. At least one **TestOutput** if the project has a test suite
-3. Do NOT output only TestOutput/TerminalRecording when the feature involves visible UI changes — screenshots are mandatory in that case
+3. Do NOT output only TestOutput/TerminalRecording when the feature involves visible UI changes — app-level screenshots are mandatory in that case
+4. Do NOT submit only Storybook screenshots for UI features — each UI screenshot MUST have a corresponding app-level screenshot proving the feature works in its real context
 
 ${storageSection}
 

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/prompts/fast-implement.prompt.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"fast-implement.prompt.d.ts","sourceRoot":"","sources":["../../../../../../../../../../packages/core/src/infrastructure/services/agents/feature-agent/nodes/prompts/fast-implement.prompt.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAMH,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,gBAAgB,CAAC;AAwIxD;;;;;;;;;;;GAWG;AACH,wBAAgB,wBAAwB,CAAC,KAAK,EAAE,iBAAiB,GAAG,MAAM,CAuEzE"}
+{"version":3,"file":"fast-implement.prompt.d.ts","sourceRoot":"","sources":["../../../../../../../../../../packages/core/src/infrastructure/services/agents/feature-agent/nodes/prompts/fast-implement.prompt.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAMH,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,gBAAgB,CAAC;AAwIxD;;;;;;;;;;;GAWG;AACH,wBAAgB,wBAAwB,CAAC,KAAK,EAAE,iBAAiB,GAAG,MAAM,CAoEzE"}

package/dist/packages/core/src/infrastructure/services/agents/feature-agent/nodes/prompts/fast-implement.prompt.js

@@ -211,9 +211,6 @@ ${dirListing}
 - Follow existing codebase conventions and patterns
 - Do NOT commit, push, or create pull requests
 - Do NOT modify any spec YAML files
-- Keep changes focused and minimal
-- Do NOT enter plan mode — implement directly without planning phases
-- Do NOT ask the user questions or use AskUserQuestion — make reasonable decisions and proceed
-- You MUST create or modify actual code files — a plan or summary alone is not acceptable output`);
+- Keep changes focused and minimal`);
     return sections.join('\n\n');
 }