@colin4k1024/tsp 2.4.1 → 2.4.2

package/docs/presentation/generate_capability_matrix.py

@@ -0,0 +1,396 @@
+#!/usr/bin/env python3
+"""Generate Agent Capability Matrix PPTX - Updated with new capabilities"""
+
+from pptx import Presentation
+from pptx.util import Inches, Pt
+from pptx.enum.text import PP_ALIGN
+
+prs = Presentation()
+prs.slide_width = Inches(10)
+prs.slide_height = Inches(7.5)
+
+def add_slide(prs):
+    return prs.slides.add_slide(prs.slide_layouts[6])
+
+def set_title(slide, text, top=0.4):
+    txBox = slide.shapes.add_textbox(Inches(0.5), Inches(top), Inches(9), Inches(0.8))
+    tf = txBox.text_frame
+    p = tf.paragraphs[0]
+    p.text = text
+    p.font.size = Pt(32)
+    p.font.bold = True
+    return txBox
+
+def add_bullets(slide, bullets, left=0.5, top=1.4, width=9, font_size=18, line_spacing=24):
+    txBox = slide.shapes.add_textbox(Inches(left), Inches(top), Inches(width), Inches(5.5))
+    tf = txBox.text_frame
+    tf.word_wrap = True
+    for i, b in enumerate(bullets):
+        p = tf.paragraphs[0] if i == 0 else tf.add_paragraph()
+        p.text = f'• {b}'
+        p.font.size = Pt(font_size)
+        p.space_after = Pt(line_spacing - font_size)
+    return txBox
+
+def add_two_columns(slide, title, left_title, left_items, right_title, right_items, top=0.4):
+    set_title(slide, title, top)
+    lt_box = slide.shapes.add_textbox(Inches(0.5), Inches(1.3), Inches(4.3), Inches(0.5))
+    lt = lt_box.text_frame.paragraphs[0]
+    lt.text = left_title
+    lt.font.size = Pt(20)
+    lt.font.bold = True
+    lb = slide.shapes.add_textbox(Inches(0.5), Inches(1.8), Inches(4.3), Inches(5))
+    ltf = lb.text_frame
+    ltf.word_wrap = True
+    for i, b in enumerate(left_items):
+        p = ltf.paragraphs[0] if i == 0 else ltf.add_paragraph()
+        p.text = f'• {b}'
+        p.font.size = Pt(16)
+        p.space_after = Pt(8)
+    rt_box = slide.shapes.add_textbox(Inches(5.2), Inches(1.3), Inches(4.3), Inches(0.5))
+    rt = rt_box.text_frame.paragraphs[0]
+    rt.text = right_title
+    rt.font.size = Pt(20)
+    rt.font.bold = True
+    rb = slide.shapes.add_textbox(Inches(5.2), Inches(1.8), Inches(4.3), Inches(5))
+    rtf = rb.text_frame
+    rtf.word_wrap = True
+    for i, b in enumerate(right_items):
+        p = rtf.paragraphs[0] if i == 0 else rtf.add_paragraph()
+        p.text = f'• {b}'
+        p.font.size = Pt(16)
+        p.space_after = Pt(8)
+
+# Slide 1: Cover
+s1 = add_slide(prs)
+txBox = s1.shapes.add_textbox(Inches(0.5), Inches(2.8), Inches(9), Inches(1.5))
+tf = txBox.text_frame
+p = tf.paragraphs[0]
+p.text = 'Agent 能力矩阵'
+p.font.size = Pt(48)
+p.font.bold = True
+p.alignment = PP_ALIGN.CENTER
+txBox2 = s1.shapes.add_textbox(Inches(0.5), Inches(4.3), Inches(9), Inches(1))
+tf2 = txBox2.text_frame
+p2 = tf2.paragraphs[0]
+p2.text = 'Harness Layer 现状分析与发展规划'
+p2.font.size = Pt(24)
+p2.alignment = PP_ALIGN.CENTER
+txBox3 = s1.shapes.add_textbox(Inches(0.5), Inches(5.3), Inches(9), Inches(0.5))
+tf3 = txBox3.text_frame
+p3 = tf3.paragraphs[0]
+p3.text = '2026-03-29'
+p3.font.size = Pt(16)
+p3.alignment = PP_ALIGN.CENTER
+
+# Slide 2: 能力分类概览
+s2 = add_slide(prs)
+set_title(s2, 'Agent 能力体系分类')
+add_bullets(s2, [
+    'Token/Context Management — 上下文窗口管理、模型选择、令牌优化',
+    'Memory & Persistence — 跨会话记忆、自动保存/加载、上下文恢复',
+    'Learning & Adaptation — 从会话中提取模式、持续优化、错误经验沉淀',
+    'Verification & Evaluation — 检查点验证、评分器、pass@k 指标',
+    'Parallelization & Scaling — 多实例并行、任务分发、动态扩展',
+    'Sub-agent Orchestration — 多代理协作、上下文传递、迭代检索'
+], font_size=20, line_spacing=32)
+
+# Slide 3: 你填写的 6 项能力
+s3 = add_slide(prs)
+set_title(s3, '你已定义的能力范畴')
+add_bullets(s3, [
+    'Token 优化 — 模型选择、系统提示精简、后台进程',
+    '内存持久化 — 自动跨会话保存/加载上下文的钩子',
+    '持续学习 — 从会话中自动提取模式到可重用的技能',
+    '验证循环 — 检查点 vs 持续评估、评分器类型、pass@k 指标',
+    '并行化 — Git worktrees、级联方法、何时扩展实例',
+    '子代理编排 — 上下文问题、迭代检索模式'
+], font_size=20, line_spacing=32)
+
+# Slide 4: 能力矩阵现状（最新）
+s4 = add_slide(prs)
+set_title(s4, '能力矩阵现状分析（最新）')
+rows = [
+    ['能力领域', '当前状态', '实现方式'],
+    ['Token/Context Management', '✅ 已实现', 'pre_compact + suggest_compact + trigger_compact'],
+    ['Memory & Persistence', '✅ 已实现', 'session_start/end hooks + memory_store'],
+    ['Learning & Adaptation', '✅ 已实现', 'instinct_store + observe hook + instinct_cli'],
+    ['Verification & Evaluation', '✅ 强覆盖', 'eval-harness skill + pass@k 指标'],
+    ['Parallelization & Scaling', '✅ 已实现', 'parallel-execution skill + worktree'],
+    ['Sub-agent Orchestration', '✅ 强覆盖', '8 角色矩阵、/team-* 流程链'],
+    ['Cost-Aware Pipeline', '✅ 已实现', 'cost_tracker + context_budget + Haiku/Sonnet/Opus'],
+    ['Error Handling & Recovery', '✅ 强覆盖', 'build-error-resolver + 7 语言版本'],
+    ['Security & Governance', '✅ 强覆盖', 'security-reviewer + OWASP Top 10 + 30+ 检查项'],
+    ['PreToolUse Hooks', '✅ 已实现', 'tmux/git/console/doc/mcp-health 五种钩子'],
+    ['PostToolUse Hooks', '✅ 已实现', 'quality-gate + format + pr-created 三种钩子'],
+]
+col_widths = [Inches(2.8), Inches(1.4), Inches(5.3)]
+row_height = Inches(0.55)
+start_top = Inches(1.3)
+left = Inches(0.5)
+for i, h in enumerate(rows[0]):
+    hdr = s4.shapes.add_textbox(left, start_top, col_widths[i], row_height)
+    tf = hdr.text_frame
+    p = tf.paragraphs[0]
+    p.text = h
+    p.font.size = Pt(14)
+    p.font.bold = True
+    left += col_widths[i]
+top = start_top + row_height
+for row in rows[1:]:
+    left = Inches(0.5)
+    for i, cell in enumerate(row):
+        cell_box = s4.shapes.add_textbox(left, top, col_widths[i], row_height)
+        tf = cell_box.text_frame
+        tf.word_wrap = True
+        p = tf.paragraphs[0]
+        p.text = cell
+        p.font.size = Pt(13)
+        left += col_widths[i]
+    top += row_height
+
+# Slide 5: Memory & Persistence 详情
+s5 = add_slide(prs)
+add_two_columns(s5, 'Memory & Persistence 已实现',
+    'session_start hook',
+    [
+        '加载上次会话摘要',
+        '检索 pending items',
+        '检查待确认决策',
+        '提供下次会话提示',
+        '存储：~/.claude/memory/sessions/'
+    ],
+    'session_end hook',
+    [
+        '收集会话数据',
+        '保存任务列表',
+        '持久化决策快照',
+        '记录 key_findings',
+        '自动生成 session_id'
+    ]
+)
+
+# Slide 6: Context Compression 详情
+s6 = add_slide(prs)
+set_title(s6, 'Context Compression 已实现')
+add_bullets(s6, [
+    'pre_compact hook — 压缩前分析，分类 keep/summarize/discard',
+    '  └ 高价值：decision、conclusion、pending_item、verification_result',
+    '  └ 低价值：tool_result、search_result（>500字符时丢弃）',
+    'suggest_compact hook — 上下文使用 >70% 时触发压缩建议',
+    '  └ 阈值：low(<70%) / medium(70-85%) / high(85-95%) / critical(>95%)',
+    '  └ 4阶段重组计划：保存决策→保存待办→压缩对话→精简工具输出'
+], font_size=17, line_spacing=24)
+
+# Slide 7: Error Experience Library
+s7 = add_slide(prs)
+add_two_columns(s7, 'Learning & Adaptation: Error Experience Library',
+    '核心操作',
+    [
+        'record: 记录错误模式',
+        'search: 查询历史模式',
+        'feedback: 更新成功率',
+        '按 (success - failure) 排序',
+        '高成功率优先推荐'
+    ],
+    '存储结构',
+    [
+        '~/.claude/memory/error_experience/',
+        'patterns/: 错误模式库',
+        '  - error_type',
+        '  - root_cause + solution',
+        '  - success_count / failure_count',
+        'decisions/: 决策快照'
+    ]
+)
+
+# Slide 8: Parallel Execution
+s8 = add_slide(prs)
+add_two_columns(s8, 'Parallelization & Scaling 已实现',
+    '任务类型',
+    [
+        'independent: 独立任务 → 并行',
+        'sequential: 串行依赖',
+        'parallel: 多独立任务 → 并行',
+        'cascade: 主任务触发子任务'
+    ],
+    'Git Worktree 并行',
+    [
+        'worktree add <path> <branch>',
+        'worktree list',
+        'worktree prune',
+        'Scale Out: 增加实例数',
+        'Scale Up: 提升实例规格'
+    ]
+)
+
+# Slide 9: 项目已有强项
+s9 = add_slide(prs)
+add_two_columns(s9, '项目已有强项',
+    'Verification & Evaluation',
+    [
+        'maven-qa (4阶段质量门禁)',
+        'pairwise-test-design',
+        'browser-smoke-testing',
+        'testcontainers-integration',
+        'tdd-guide + e2e-runner',
+        '45+ 质量门禁检查项'
+    ],
+    'Sub-agent Orchestration',
+    [
+        '8 角色矩阵 (tech-lead 统筹)',
+        '/team-* 完整流程链',
+        '/handoff 标准化交接',
+        'chief-of-staff 跨角色协调',
+        '27 个 Specialist Agents',
+        '27 个 ECC 技能'
+    ]
+)
+
+# Slide 10: 新增能力文件清单
+s10 = add_slide(prs)
+set_title(s10, '新增能力文件清单')
+add_bullets(s10, [
+    'scripts/lib/memory_store.py — 底层存储接口',
+    'scripts/lib/instinct_store.py — Instinct 持久化（项目/全局作用域）',
+    'scripts/lib/cost_tracker.py — Token 成本追踪',
+    'scripts/lib/context_budget.py — 上下文预算监控',
+    'scripts/lib/trigger_compact.py — 触发式压缩（4种触发器）',
+    'scripts/lib/context_priority.py — 优先级分类（高/中/低）',
+    'scripts/lib/context_archiver.py — 上下文归档到内存',
+    'scripts/hooks/pre_compact.py — 增强：trigger-based 压缩',
+    'scripts/hooks/cost_tracker.py — Stop 事件成本记录',
+    'scripts/hooks/observe.py — 工具调用观察捕获',
+    'skills/eval-harness/ — Eval 驱动开发 + pass@k',
+    'skills/continuous-learning/ — Instinct 持续学习 v2.1',
+    'skills/strategic-compact/ — 上下文优化技能',
+    'skills/cost-aware-llm-pipeline/ — 成本感知 LLM 管道',
+    'commands/harness-audit.md — 7 维度评分审计命令',
+    'docs/runbooks/command-and-capability-matrix.md — 命令与能力矩阵',
+    'docs/runbooks/runtime-capabilities-overview.md — runtime 总览',
+    'docs/runbooks/platform-capability-demo-script.md — 平台升级演示剧本'
+], font_size=14, line_spacing=20)
+
+# Slide 11: PreToolUse/PostToolUse Hooks
+s11 = add_slide(prs)
+add_two_columns(s11, 'Hooks 系统增强',
+    'PreToolUse Hooks (5个)',
+    [
+        'tmux-reminder — 检测 dev server 建议用 tmux',
+        'git-push-reminder — push 前检查未提交变更',
+        'console-log-warning — 检测 console.log 等调试语句',
+        'doc-file-warning — 警告非标准位置文档',
+        'mcp-health-check — MCP 服务器健康检查'
+    ],
+    'PostToolUse Hooks (3个)',
+    [
+        'quality-gate — 编辑后质量检查（空格/分号/console）',
+        'format-suggestion — JS/TS 文件格式化建议',
+        'pr-created — gh pr create 后提取 PR URL'
+    ]
+)
+
+# Slide 12: Security Reviewer OWASP Top 10
+s12 = add_slide(prs)
+set_title(s12, 'Security Reviewer OWASP Top 10 覆盖')
+add_bullets(s12, [
+    'A01: Broken Access Control — 访问控制检查',
+    'A02: Cryptographic Failures — 加密失效检查',
+    'A03: Injection — 注入漏洞检查',
+    'A04: Insecure Design — 不安全设计检查',
+    'A05: Security Misconfiguration — 安全配置错误',
+    'A06: Vulnerable Components — 漏洞组件检查',
+    'A07: Authentication Failures — 认证失败检查',
+    'A08: Data Integrity Failures — 数据完整性检查',
+    'A09: Logging Failures — 日志记录失败检查',
+    'A10: SSRF — 服务端请求伪造检查'
+], font_size=17, line_spacing=22)
+
+# Slide 13: 架构演进路线（最新）
+s13 = add_slide(prs)
+add_two_columns(s13, '架构演进路线（最新）',
+    'Phase 1-3: 已完成 ✅',
+    [
+        'Memory Persistence hooks',
+        'Context Compression hooks',
+        'Error Experience Library',
+        'Parallel Execution Framework',
+        'PreToolUse/PostToolUse Hooks',
+        'Continuous Learning System',
+        'Cost Tracking + Context Budget',
+        'Strategic Compact 增强'
+    ],
+    'Phase 4-7: 已完成 ✅',
+    [
+        'Eval Harness + pass@k 指标',
+        'Cost-Aware LLM Pipeline',
+        'Harness Audit 7维度评分',
+        'Security Reviewer OWASP Top 10',
+        'Instinct-based Learning v2.1',
+        'Trigger-based Lazy Loading',
+        'context_archiver 上下文归档'
+    ]
+)
+
+# Slide 14: 新增命令与 runtime 边界
+s14 = add_slide(prs)
+add_two_columns(s14, '新增命令与 Runtime 边界',
+    '显式命令',
+    [
+        '/tdd — 先定义测试与完成标准',
+        '/harness-audit — 平台覆盖度体检',
+        '/team-* — 主链角色协作入口',
+        'specialist 结果必须回到 handoff 或主链'
+    ],
+    'Runtime 机制',
+    [
+        'memory persistence — 会话摘要与待办回存',
+        'observe — 工具调用模式观察',
+        'cost-aware pipeline — 成本与预算控制',
+        'compact/archive — 长会话上下文重组'
+    ]
+)
+
+# Slide 15: 演示材料与建议阅读路径
+s15 = add_slide(prs)
+set_title(s15, '演示材料与建议阅读路径')
+add_bullets(s15, [
+    '首次介绍平台：team-skills-platform-intro.pptx',
+    '能力矩阵说明：agent-capability-matrix.pptx',
+    '命令全景：command-and-capability-matrix.md',
+    'runtime 总览：runtime-capabilities-overview.md',
+    '平台升级演示：platform-capability-demo-script.md',
+    'vertical 场景导览：docs/presentation/vertical-scenario-route-map.md',
+    'vertical 场景矩阵：docs/runbooks/vertical-scenario-capability-matrix.md',
+    '讲解提词稿：docs/presentation/presentation-talk-track.md'
+], font_size=18, line_spacing=24)
+
+# Slide 16: Vertical 场景成熟度
+s16 = add_slide(prs)
+add_two_columns(s16, 'vertical 场景成熟度与使用方式',
+    '完整 demo 闭环',
+    [
+        'GitHub Actions / 供应链治理',
+        'AI / Eval 平台',
+        '移动端 / 小程序交付',
+        'IaC / Kubernetes 平台仓库',
+        '插件 / 扩展仓库',
+        '数据 / ML pipeline 仓库',
+        '安全 / 合规平台仓库',
+        '内部开发者平台',
+        '数据可观测性 / 质量平台'
+    ],
+    '建议入口',
+    [
+        '先看 vertical-scenario-route-map.md',
+        '再看 vertical-scenario-capability-matrix.md',
+        '再按项目类型选对应 demo script',
+        '需要复盘时搭配 execution log',
+        '需要接入时回到 examples/INDEX.md'
+    ]
+)
+
+import os
+output = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'agent-capability-matrix.pptx')
+prs.save(output)
+print(f'PPTX saved: {output}')