speccore 8.0.1 → 8.1.0

package/dist/commands/analyze.js

@@ -3234,6 +3234,7 @@ sequenceDiagram
     // v8.0.0+: 骨架进度检测 — 每次只生成下一个未填充的文档
     // 替代 v7.5.0 的 detectIterationDocsStatus（基于文件是否存在）
     let perDocStatus = null;
+    let lastDocQuality = null;
     if (!isGlobal && !isTask && !ctx.phase && ctx.iteration && ctx.iteration !== 'GLOBAL') {
         try {
             const iterDir = await (0, context_1.getIterationDir)(ctx.iteration);
@@ -3256,6 +3257,27 @@ sequenceDiagram
                         taskDocs = taskDocs.filter(([n]) => n === nextDocName);
                     }
                     logger_1.logger.info(`🦴 骨架进度: ${perDocStatus.filledCount}/${perDocStatus.totalCount} 已填充，下一个: ${perDocStatus.nextUnfilled.relPath}`);
+                    // v8.1.0+: 校验上一个已填充文档的质量
+                    if (perDocStatus.filled.length > 0) {
+                        const lastFilled = perDocStatus.filled[perDocStatus.filled.length - 1];
+                        const lastFilledPath = (0, path_1.join)(specDir, lastFilled);
+                        if (await (0, fs_extra_1.pathExists)(lastFilledPath)) {
+                            try {
+                                const lastContent = await (0, fs_extra_2.readFile)(lastFilledPath, 'utf-8');
+                                const lastDocName = lastFilled.includes('/') ? lastFilled.split('/').pop() : lastFilled;
+                                const lastPlatform = lastFilled.includes('/') ? lastFilled.split('/')[0] : undefined;
+                                const quality = await (0, spec_skeleton_1.validateContentQuality)(lastDocName, lastContent, lastPlatform !== lastDocName ? lastPlatform : undefined);
+                                lastDocQuality = { docName: lastFilled, score: quality.score, issues: quality.issues };
+                                if (quality.score < 60) {
+                                    logger_1.logger.warn(`⚠️ ${lastFilled} 质量评分: ${quality.score}/100 — ${quality.issues.join('；')}`);
+                                }
+                                else if (quality.score < 80) {
+                                    logger_1.logger.info(`📝 ${lastFilled} 质量评分: ${quality.score}/100 — ${quality.issues.length > 0 ? quality.issues.join('；') : '良好'}`);
+                                }
+                            }
+                            catch { /* skip */ }
+                        }
+                    }
                 }
                 else if (perDocStatus.filledCount === perDocStatus.totalCount) {
                     logger_1.logger.info(`🎉 所有 ${perDocStatus.totalCount} 个骨架文件均已填充`);
@@ -3453,6 +3475,53 @@ sequenceDiagram
     prompt += `- 只覆盖上述已存在的文件，不要创建新文件或新目录\n`;
     prompt += `- 每个文件写入正确路径（如 \`overview/ANALYSIS.md\`，不是根目录的 \`ANALYSIS.md\`）\n`;
     prompt += `- 写完后用 Read 验证文件内容已替换占位内容\n\n`;
+    // ── v8.1.0+: 逐文档模式自动注入 PRD 内容 + 前序文档摘要 ──
+    if (perDocStatus?.nextUnfilled) {
+        try {
+            const iterDirForCtx = await (0, context_1.getIterationDir)(iter);
+            if (iterDirForCtx) {
+                // 1. 注入 PRD 内容（010-requirements/ 下的需求文档）
+                const reqDir = (0, path_1.join)(iterDirForCtx, '010-requirements');
+                let prdContent = '';
+                const prdSources = (0, path_1.join)(reqDir, 'sources');
+                const prdConverted = (0, path_1.join)(reqDir, 'converted');
+                const prdFeatures = (0, path_1.join)(reqDir, 'features');
+                for (const dir of [prdSources, prdConverted, prdFeatures]) {
+                    if (await (0, fs_extra_1.pathExists)(dir)) {
+                        const files = (await Promise.resolve().then(() => __importStar(require('fs-extra')))).readdirSync(dir).filter((f) => f.endsWith('.md')).slice(0, 5);
+                        for (const f of files) {
+                            const content = await (await Promise.resolve().then(() => __importStar(require('fs-extra')))).readFile((0, path_1.join)(dir, f), 'utf-8');
+                            prdContent += `\n### ${f}\n${content.slice(0, 2000)}\n`;
+                        }
+                    }
+                }
+                if (prdContent.length > 0) {
+                    prompt += `\n## 📎 需求文档原文（PRD，写作核心输入）\n\n`;
+                    prompt += `> 以下是 010-requirements/ 下的需求文档，这是你撰写专业分析文档的核心输入。所有分析必须基于这些内容，不要臆造。\n`;
+                    prompt += prdContent.slice(0, 5000);
+                    prompt += `\n\n`;
+                }
+                // 2. 注入前序已填充文档摘要
+                const specDir = (0, path_1.join)(iterDirForCtx, '020-specs');
+                if (perDocStatus.filled.length > 0) {
+                    prompt += `## 📎 前序已填充文档摘要（当前文档必须与之保持一致）\n\n`;
+                    prompt += `> 以下文档已生成，当前文档必须与之保持一致（字段名、接口路径、状态枚举等不能冲突）。\n\n`;
+                    for (const filledPath of perDocStatus.filled) {
+                        const fullPath = (0, path_1.join)(specDir, filledPath);
+                        try {
+                            const content = await (await Promise.resolve().then(() => __importStar(require('fs-extra')))).readFile(fullPath, 'utf-8');
+                            const summary = content.slice(0, 1500);
+                            prompt += `### ${filledPath}\n\`\`\`\n${summary}${content.length > 1500 ? '\n... (截断)' : ''}\n\`\`\`\n\n`;
+                        }
+                        catch { /* skip unreadable */ }
+                    }
+                }
+            }
+        }
+        catch (e) {
+            logger_1.logger.debug?.(`PRD/前序文档注入失败（非关键）: ${e.message}`);
+        }
+    }
     prompt += `## ⚠️ 迭代名称仅为目录标识（重要）\n\n`;
     prompt += `- 迭代名称（"${iter}"）仅为目录标识符，**不代表需求内容**\n`;
     prompt += `- **需求文档是唯一事实来源**：所有分析必须 100% 基于 010-requirements/ 下的文档内容\n`;
@@ -3802,6 +3871,12 @@ sequenceDiagram
         prompt += `> 当前为**单文档模式**：本次只生成 1 个文档，写入后立即推进下一个。\n`;
         prompt += `> 骨架进度: ${perDocStatus.filledCount}/${perDocStatus.totalCount} 已填充，剩余 ${perDocStatus.unfilled.length} 个\n`;
         prompt += `> 待填充: ${perDocStatus.unfilled.slice(0, 5).join('、')}${perDocStatus.unfilled.length > 5 ? '...' : ''}\n\n`;
+        // v8.1.0+: 注入上一文档质量警告
+        if (lastDocQuality && lastDocQuality.score < 80) {
+            prompt += `> ⚠️ **上一文档 ${lastDocQuality.docName} 质量评分仅 ${lastDocQuality.score}/100**\n`;
+            prompt += `> 问题: ${lastDocQuality.issues.join('；')}\n`;
+            prompt += `> 请确保当前文档达到专业标准——内容充实、章节完整、无占位符残留\n\n`;
+        }
         if (remainingAfterThis > 0) {
             prompt += `**当前文档写入完成后，立即输出以下标记推进下一个文档**：\n\n`;
             prompt += `[SPECCORE_EXEC: speccore analyze --prompt -I ${iter}${options?.auto ? ' --auto' : ''}${options?.withCode ? ' --with-code' : ''}]\n\n`;