mstro-app 0.4.20 → 0.4.22

package/server/services/plan/review-gate.ts

@@ -13,6 +13,7 @@ import { existsSync, mkdirSync, readdirSync, readFileSync, writeFileSync } from
 import { join } from 'node:path';
 import { runWithFileLogger } from '../../cli/headless/headless-logger.js';
 import { HeadlessRunner } from '../../cli/headless/index.js';
+import { loadAgentPrompt } from './agent-loader.js';
 import type { Issue, ReviewCheck, ReviewResult } from './types.js';
 
 /** Max review attempts per issue per sprint before giving up */
@@ -33,6 +34,10 @@ export interface ReviewIssueOptions {
   logDir?: string;
   /** Custom board-level review criteria — replaces default review instructions when set */
   reviewCriteria?: string;
+  /** Board directory for agent prompt override resolution (e.g., .mstro/pm/boards/BOARD-001) */
+  boardDir?: string | null;
+  /** Extra environment variables for spawned Claude processes (e.g. API keys) */
+  extraEnv?: Record<string, string>;
 }
 
 /**
@@ -40,12 +45,12 @@ export interface ReviewIssueOptions {
  * Returns auto-pass on infrastructure failures to avoid blocking execution.
  */
 export async function reviewIssue(options: ReviewIssueOptions): Promise<ReviewResult> {
-  const { workingDir, issue, pmDir, outputPath, onOutput, logDir, reviewCriteria } = options;
+  const { workingDir, issue, pmDir, outputPath, onOutput, logDir, reviewCriteria, boardDir } = options;
   const isCodeTask = issue.filesToModify.length > 0;
   const issueType: ReviewResult['issueType'] = isCodeTask ? 'code' : 'non-code';
 
   try {
-    const prompt = buildReviewPrompt(issue, pmDir, outputPath, isCodeTask, reviewCriteria);
+    const prompt = buildReviewPrompt(issue, pmDir, outputPath, isCodeTask, reviewCriteria, boardDir);
 
     const runner = new HeadlessRunner({
       workingDir,
@@ -53,8 +58,9 @@ export async function reviewIssue(options: ReviewIssueOptions): Promise<ReviewRe
       stallWarningMs: REVIEW_STALL_WARNING_MS,
       stallKillMs: REVIEW_STALL_KILL_MS,
       stallHardCapMs: REVIEW_STALL_HARD_CAP_MS,
-      verbose: false,
+      verbose: true,
       outputCallback: onOutput ? (text: string) => onOutput(`Review: ${text}`) : undefined,
+      extraEnv: options.extraEnv,
     });
 
     const result = await runWithFileLogger('pm-review', () => runner.run(), logDir);
@@ -175,74 +181,139 @@ export function autoPassResult(issueId: string, issueType: ReviewResult['issueTy
 
 // ── Private helpers ─────────────────────────────────────────
 
-function buildReviewPrompt(issue: Issue, pmDir: string, outputPath: string, isCodeTask: boolean, reviewCriteria?: string): string {
+function buildReviewPrompt(
+  issue: Issue,
+  pmDir: string,
+  outputPath: string,
+  isCodeTask: boolean,
+  reviewCriteria?: string,
+  boardDir?: string | null,
+): string {
   const criteria = issue.acceptanceCriteria
     .map(c => `- [${c.checked ? 'x' : ' '}] ${c.text}`)
     .join('\n');
+  const criteriaStr = criteria || 'No specific criteria defined.';
+  const filesModified = issue.filesToModify.map(f => `- ${f}`).join('\n');
+  const issueSpecPath = join(pmDir, issue.path);
 
-  // When custom review criteria are set, use a generic review prompt
-  // that applies the user's criteria instead of assuming code review.
+  // When custom review criteria are set, use the review-custom agent.
   if (reviewCriteria) {
-    return `You are a reviewer. Review the work done for issue ${issue.id}: ${issue.title}.
-${isCodeTask ? `\n## Files Modified\n${issue.filesToModify.map(f => `- ${f}`).join('\n')}` : `\n## Output File\n${outputPath}\n\n## Issue Spec\n${join(pmDir, issue.path)}`}
-
-## Acceptance Criteria
-${criteria || 'No specific criteria defined.'}
-
-## Review Criteria
-${reviewCriteria}
+    const contextSection = isCodeTask
+      ? `\n## Files Modified\n${filesModified}`
+      : `\n## Output File\n${outputPath}\n\n## Issue Spec\n${issueSpecPath}`;
+    const readInstruction = isCodeTask
+      ? 'Read each modified file listed above'
+      : 'Read the output file and issue spec at the paths above';
+
+    return loadAgentPrompt('review-custom', {
+      issue_id: issue.id,
+      issue_title: issue.title,
+      context_section: contextSection,
+      acceptance_criteria: criteriaStr,
+      review_criteria: reviewCriteria,
+      read_instruction: readInstruction,
+    }, boardDir) ?? buildCustomFallback(issue, contextSection, criteriaStr, reviewCriteria, readInstruction);
+  }
 
-## Instructions
-1. ${isCodeTask ? 'Read each modified file listed above' : 'Read the output file and issue spec at the paths above'}
-2. Check if all acceptance criteria are met
-3. Evaluate against the review criteria above
+  if (isCodeTask) {
+    return loadAgentPrompt('review-code', {
+      issue_id: issue.id,
+      issue_title: issue.title,
+      files_modified: filesModified,
+      acceptance_criteria: criteriaStr,
+      output_path: outputPath,
+    }, boardDir) ?? buildCodeFallback(issue, filesModified, criteriaStr, outputPath);
+  }
 
-Output EXACTLY one JSON object on its own line (no markdown fencing):
-{"passed": true, "checks": [{"name": "criteria_met", "passed": true, "details": "..."}]}
+  return loadAgentPrompt('review-quality', {
+    issue_id: issue.id,
+    issue_title: issue.title,
+    output_path: outputPath,
+    issue_spec_path: issueSpecPath,
+    acceptance_criteria: criteriaStr,
+  }, boardDir) ?? buildQualityFallback(issue, outputPath, issueSpecPath, criteriaStr);
+}
 
-Include checks for: criteria_met, review_criteria.`;
-  }
+// ── Hardcoded fallbacks (used when agent files are missing) ──────────
 
-  if (isCodeTask) {
-    return `You are a code reviewer. Review the work done for issue ${issue.id}: ${issue.title}.
+function buildCodeFallback(issue: Issue, filesModified: string, criteria: string, outputPath: string): string {
+  return `You are a reviewer. Review the work done for issue ${issue.id}: ${issue.title}.
 
 ## Files Modified
-${issue.filesToModify.map(f => `- ${f}`).join('\n')}
+${filesModified}
 
 ## Acceptance Criteria
-${criteria || 'No specific criteria defined.'}
+${criteria}
 
 ## Instructions
 1. Read each modified file listed above
-2. Check if all acceptance criteria are met by the code changes
-3. Look for obvious bugs, security vulnerabilities, or code quality issues
+2. Check if all acceptance criteria are met by the changes
+3. Evaluate the quality of the changes:
+   - For source code files: look for obvious bugs, security vulnerabilities, or code quality issues
+   - For content files (markdown, docs, config, copy): check for accuracy, completeness, and appropriate structure
 4. Check if the output artifact exists at: ${outputPath}
 
 Output EXACTLY one JSON object on its own line (no markdown fencing):
 {"passed": true, "checks": [{"name": "criteria_met", "passed": true, "details": "..."}]}
 
 Include checks for: criteria_met, code_quality, no_obvious_bugs.`;
-  }
+}
 
+function buildQualityFallback(issue: Issue, outputPath: string, issueSpecPath: string, criteria: string): string {
   return `You are a quality reviewer. Review the work done for issue ${issue.id}: ${issue.title}.
 
 ## Output File
 ${outputPath}
 
 ## Issue Spec
-${join(pmDir, issue.path)}
+${issueSpecPath}
 
 ## Acceptance Criteria
-${criteria || 'No specific criteria defined.'}
+${criteria}
 
 ## Instructions
 1. Read the output file at the path above
-2. Read the full issue spec
-3. Check if all acceptance criteria are met
-4. Check for completeness and quality of the output
+2. Read the full issue spec to understand the original requirements and intent
+3. Evaluate the output against ALL of the following dimensions:
+
+### Acceptance Criteria
+- Are all acceptance criteria met? Check each one individually.
+
+### Content Quality
+- Is the content accurate, well-reasoned, and free of factual errors?
+- Is it written clearly with appropriate structure and organization?
+- Does it have sufficient depth and detail for its purpose?
+- Is the tone and style appropriate for the intended audience?
+
+### Completeness
+- Does the output fully address what was requested in the issue spec?
+- Are there obvious gaps, missing sections, or incomplete thoughts?
+- If the issue requested specific deliverables (e.g., a plan, analysis, document), are all deliverables present?
 
 Output EXACTLY one JSON object on its own line (no markdown fencing):
 {"passed": true, "checks": [{"name": "criteria_met", "passed": true, "details": "..."}]}
 
 Include checks for: criteria_met, output_quality, completeness.`;
 }
+
+function buildCustomFallback(issue: Issue, contextSection: string, criteria: string, reviewCriteria: string, readInstruction: string): string {
+  return `You are a reviewer. Review the work done for issue ${issue.id}: ${issue.title}.
+${contextSection}
+
+## Acceptance Criteria
+${criteria}
+
+## Review Criteria
+${reviewCriteria}
+
+## Instructions
+1. ${readInstruction}
+2. Check if all acceptance criteria are met — evaluate each criterion individually
+3. Evaluate thoroughly against the review criteria above
+4. Consider the overall quality of the work: does it fully address the issue's intent, is it well-structured, and is it ready to ship?
+
+Output EXACTLY one JSON object on its own line (no markdown fencing):
+{"passed": true, "checks": [{"name": "criteria_met", "passed": true, "details": "..."}]}
+
+Include checks for: criteria_met, review_criteria.`;
+}

package/server/services/plan/state-reconciler.ts

@@ -111,27 +111,36 @@ function buildStateMarkdown(
 
 /**
  * Derive epic status from its children's actual statuses.
- * All children done/cancelled → done (auto-complete the epic).
+ * - All children done/cancelled → done
+ * - Any child in_progress/in_review → in_progress
+ * - Otherwise → null (no change)
  */
-function deriveEpicDone(epic: Issue, issueByPath: Map<string, Issue>): boolean {
-  if (epic.children.length === 0) return false;
-  if (epic.status === 'done' || epic.status === 'cancelled') return false;
+function deriveEpicStatus(epic: Issue, issueByPath: Map<string, Issue>): string | null {
+  if (epic.children.length === 0) return null;
+  if (epic.status === 'done' || epic.status === 'cancelled') return null;
 
-  return epic.children.every(childPath => {
-    const child = issueByPath.get(childPath);
-    return child && (child.status === 'done' || child.status === 'cancelled');
-  });
+  const childStatuses = epic.children.map(cp => issueByPath.get(cp)?.status).filter(Boolean) as string[];
+  if (childStatuses.length === 0) return null;
+
+  const allFinished = childStatuses.every(s => s === 'done' || s === 'cancelled');
+  if (allFinished) return 'done';
+
+  const anyStarted = childStatuses.some(s => s === 'in_progress' || s === 'in_review');
+  if (anyStarted && epic.status !== 'in_progress') return 'in_progress';
+
+  return null;
 }
 
 function reconcileEpicStatuses(pmDir: string, issues: Issue[], issueByPath: Map<string, Issue>): void {
   const epics = issues.filter(i => i.type === 'epic');
   for (const epic of epics) {
-    if (!deriveEpicDone(epic, issueByPath)) continue;
+    const derived = deriveEpicStatus(epic, issueByPath);
+    if (!derived) continue;
 
     const epicPath = join(pmDir, epic.path);
     try {
       let content = readFileSync(epicPath, 'utf-8');
-      content = replaceFrontMatterField(content, 'status', 'done');
+      content = replaceFrontMatterField(content, 'status', derived);
       writeFileSync(epicPath, content, 'utf-8');
     } catch {
       // Epic file may be missing or unwritable
@@ -206,7 +215,8 @@ export function tryCompleteParentEpic(workingDir: string, updatedIssue: Issue):
   if (!epic) return null;
 
   const issueByPath = new Map(issues.map(i => [i.path, i]));
-  if (!deriveEpicDone(epic, issueByPath)) return null;
+  const derived = deriveEpicStatus(epic, issueByPath);
+  if (derived !== 'done') return null;
 
   const epicFullPath = join(pmDir, epic.path);
   try {

package/server/services/plan/types.ts

@@ -4,7 +4,7 @@
 /**
  * Plan Types — Project Plan Spec (PPS) data structures
  *
- * These types represent the parsed contents of .pm/ directory files.
+ * These types represent the parsed contents of .mstro/pm/ directory files.
  */
 
 // ============================================================================
@@ -25,7 +25,7 @@ export interface ProjectConfig {
 
 export interface WorkflowStatus {
   status: string;
-  category: 'unstarted' | 'started' | 'completed' | 'cancelled';
+  category: 'ready' | 'unstarted' | 'started' | 'completed' | 'cancelled';
   description: string;
 }
 
@@ -98,7 +98,7 @@ export interface Issue {
   outputFile: string | null;
   // Full markdown body
   body: string;
-  // File path relative to .pm/
+  // File path relative to .mstro/pm/
   path: string;
 }
 

package/server/services/plan/watcher.ts

@@ -2,7 +2,7 @@
 // Licensed under the MIT License. See LICENSE file for details.
 
 /**
- * Plan Watcher — Watches .pm/ (or legacy .plan/) directory for changes and broadcasts updates.
+ * Plan Watcher — Watches .mstro/pm/ directory for changes and broadcasts updates.
  *
  * Uses fs.watch with debouncing to batch rapid changes.
  */

package/server/services/sentry.ts

@@ -6,9 +6,12 @@ import { homedir } from 'node:os'
 import { join } from 'node:path'
 import * as Sentry from '@sentry/node'
 
-// Hardcoded DSN for production - this is safe to expose (can only send, not read)
-// Override with SENTRY_DSN env var for development/testing
-const SENTRY_DSN = process.env.SENTRY_DSN || 'https://2a8d2493e3ee5a7beec30f4518a5e24c@o4510824844820480.ingest.us.sentry.io/4510824923594752'
+// Sentry DSN lives on the platform server. The CLI sends envelopes
+// to the server's /sentry-tunnel endpoint which proxies to Sentry.
+// A placeholder DSN is needed so the Sentry SDK initializes its
+// transport — the real DSN is injected server-side before forwarding.
+const SENTRY_TUNNEL_DSN = 'https://tunnel@sentry.io/0'
+const PLATFORM_URL = process.env.PLATFORM_URL || 'https://api.mstro.app'
 
 const CONFIG_FILE = join(homedir(), '.mstro', 'config.json')
 
@@ -51,7 +54,8 @@ export function initSentry(): void {
   initialized = true
 
   Sentry.init({
-    dsn: SENTRY_DSN,
+    dsn: SENTRY_TUNNEL_DSN,
+    tunnel: `${PLATFORM_URL}/sentry-tunnel`,
     environment: process.env.NODE_ENV || 'development',
     release: `mstro-cli@${process.env.npm_package_version || '0.0.0'}`,
     tracesSampleRate: 0.1,