@microi.net/cli 5.6.0 → 5.6.1

package/skills/scripts/validate-progressive-disclosure.mjs

@@ -1,52 +1,57 @@
-import crypto from 'node:crypto'
-import fs from 'node:fs'
-import path from 'node:path'
-import { fileURLToPath } from 'node:url'
-
-const skillsRoot = path.resolve(path.dirname(fileURLToPath(import.meta.url)), '..')
-const manifest = JSON.parse(fs.readFileSync(path.join(skillsRoot, '.progressive-disclosure-manifest.json'), 'utf8'))
-const hash = (value) => crypto.createHash('sha256').update(value).digest('hex')
-const lineCount = (value) => value.length === 0 ? 0 : value.split(/\r?\n/).length
-
-let checkedChunks = 0
-for (const [skillName, skill] of Object.entries(manifest.skills)) {
-  const entryPath = path.join(skillsRoot, skillName, 'SKILL.md')
-  const entry = fs.readFileSync(entryPath, 'utf8')
-  const begin = entry.indexOf('<!-- microi-progressive:begin -->')
-  if (begin < 0) throw new Error(`${skillName}: missing progressive begin marker`)
-  const prefix = entry.slice(0, begin)
-  if (hash(prefix) !== skill.prefixSha256) throw new Error(`${skillName}: entry prefix changed`)
-  if (lineCount(entry) > manifest.policy.maxEntryLines) throw new Error(`${skillName}: entry exceeds line budget`)
-
-  const sources = new Map([[`${skillName}/SKILL.md`, entry]])
-  for (const reference of skill.references) {
-    const relative = `${skillName}/${reference.path}`
-    const content = fs.readFileSync(path.join(skillsRoot, relative), 'utf8')
-    if (lineCount(content) > 300) throw new Error(`${relative}: reference exceeds 300 lines`)
-    sources.set(relative, content)
-  }
-
-  const ordered = []
-  for (const chunk of skill.chunks) {
-    const source = sources.get(chunk.destination)
-    if (source == null) throw new Error(`${skillName}: missing ${chunk.destination}`)
-    const open = `<!-- microi-progressive:chunk id=${chunk.id} sha256=${chunk.sha256} -->`
-    const startMarker = source.indexOf(open)
-    if (startMarker < 0) throw new Error(`${skillName}: missing chunk ${chunk.id}`)
-    const contentStart = startMarker + open.length + (source.startsWith('\r\n', startMarker + open.length) ? 2 : 1)
-    const close = '<!-- /microi-progressive:chunk -->'
-    const contentEnd = source.indexOf(close, contentStart)
-    if (contentEnd < 0) throw new Error(`${skillName}: unterminated chunk ${chunk.id}`)
-    const raw = source.slice(contentStart, contentEnd)
-    if (hash(raw) !== chunk.sha256) throw new Error(`${skillName}: chunk ${chunk.id} changed`)
-    ordered[chunk.ordinal] = raw
-    checkedChunks += 1
-  }
-  const reconstructed = prefix + ordered.join('')
-  if (hash(reconstructed) !== skill.originalSha256) {
-    throw new Error(`${skillName}: reconstructed knowledge differs from original`)
-  }
-}
-
-console.log(`Validated ${Object.keys(manifest.skills).length} skills and ${checkedChunks} lossless chunks.`)
-
+import crypto from 'node:crypto'
+import fs from 'node:fs'
+import path from 'node:path'
+import { fileURLToPath } from 'node:url'
+
+const skillsRoot = path.resolve(path.dirname(fileURLToPath(import.meta.url)), '..')
+const manifest = JSON.parse(fs.readFileSync(path.join(skillsRoot, '.progressive-disclosure-manifest.json'), 'utf8'))
+const hash = (value) => crypto.createHash('sha256').update(value).digest('hex')
+const lineCount = (value) => value.length === 0 ? 0 : value.split(/\r?\n/).length
+
+// 可只验证显式修改的 Skill；默认仍检查全库，不能为并行工作区的历史漂移改写共同基线。
+const requested = process.argv.slice(2)
+for (const name of requested) {
+  if (!Object.hasOwn(manifest.skills, name)) throw new Error(`Unknown progressive-disclosure skill: ${name}`)
+}
+const selected = Object.entries(manifest.skills).filter(([name]) => requested.length === 0 || requested.includes(name))
+let checkedChunks = 0
+for (const [skillName, skill] of selected) {
+  const entryPath = path.join(skillsRoot, skillName, 'SKILL.md')
+  const entry = fs.readFileSync(entryPath, 'utf8')
+  const begin = entry.indexOf('<!-- microi-progressive:begin -->')
+  if (begin < 0) throw new Error(`${skillName}: missing progressive begin marker`)
+  const prefix = entry.slice(0, begin)
+  if (hash(prefix) !== skill.prefixSha256) throw new Error(`${skillName}: entry prefix changed`)
+  if (lineCount(entry) > manifest.policy.maxEntryLines) throw new Error(`${skillName}: entry exceeds line budget`)
+
+  const sources = new Map([[`${skillName}/SKILL.md`, entry]])
+  for (const reference of skill.references) {
+    const relative = `${skillName}/${reference.path}`
+    const content = fs.readFileSync(path.join(skillsRoot, relative), 'utf8')
+    if (lineCount(content) > 300) throw new Error(`${relative}: reference exceeds 300 lines`)
+    sources.set(relative, content)
+  }
+
+  const ordered = []
+  for (const chunk of skill.chunks) {
+    const source = sources.get(chunk.destination)
+    if (source == null) throw new Error(`${skillName}: missing ${chunk.destination}`)
+    const open = `<!-- microi-progressive:chunk id=${chunk.id} sha256=${chunk.sha256} -->`
+    const startMarker = source.indexOf(open)
+    if (startMarker < 0) throw new Error(`${skillName}: missing chunk ${chunk.id}`)
+    const contentStart = startMarker + open.length + (source.startsWith('\r\n', startMarker + open.length) ? 2 : 1)
+    const close = '<!-- /microi-progressive:chunk -->'
+    const contentEnd = source.indexOf(close, contentStart)
+    if (contentEnd < 0) throw new Error(`${skillName}: unterminated chunk ${chunk.id}`)
+    const raw = source.slice(contentStart, contentEnd)
+    if (hash(raw) !== chunk.sha256) throw new Error(`${skillName}: chunk ${chunk.id} changed`)
+    ordered[chunk.ordinal] = raw
+    checkedChunks += 1
+  }
+  const reconstructed = prefix + ordered.join('')
+  if (hash(reconstructed) !== skill.originalSha256) {
+    throw new Error(`${skillName}: reconstructed knowledge differs from original`)
+  }
+}
+
+console.log(`Validated ${selected.length} skills and ${checkedChunks} lossless chunks${requested.length ? ' (explicit scope)' : ''}.`)

package/skills/system-observability/SKILL.md

@@ -15,16 +15,28 @@ description: Microi 系统日志/监控查询、诊断与治理规范。用于
 |---|---|
 | 人工排查、看趋势、打开日志详情 | 平台菜单【系统日志/监控】 |
 | AI 查询、自动诊断、验收 | MCP `microi_query_system_observability` |
-| AI 封禁或解封 IP | MCP `microi_manage_system_observability` |
+| AI 封禁或解封 IP | MCP `microi_manage_system_observability` |
+| 连接池耗尽时在线诊断/恢复 | 查询 `DatabasePools / DatabasePoolRecovery`，治理 `ResetDatabasePools` |
 | 应用页面读取 | Managed 接口引擎 `mci-system-observability-query` |
 | 应用页面治理 | Managed 接口引擎 `mci-system-observability-action` |
 | 扩展宿主、进程、Mongo 或安全底层原子能力 | `V8.Method.GetSystemObservability` / `V8.Method.ManageSystemObservability` |
 
 AI 第一次使用时先查询 `action=Capabilities`，再按返回的动作、权限和边界选择查询。优先使用专用工具，它已经限制动作、参数、分页、确认和审计。旧进程缺少新 Memory 动作时，仅允许通过标准 `microi_run_engine` 临时只读执行固定查询引擎；完整版本排查与降级边界见 [内存事故排查手册](references/memory-incident-triage.md)。不得创建临时维护引擎或绕过后端权限。
 
-## 查询动作
-
-`microi_query_system_observability` 支持：
+## 查询动作
+
+### 连接池故障应急入口
+
+- 收到 `obtaining a connection from the pool` 时，直接用 `microi_query_system_observability(action=DatabasePools,poolTarget=Both)`；故障池可能阻塞普通 `Capabilities` 或 Managed 引擎，此动作走固定 `/api/Diagnostics/database-pools` 应急协议。要求后端支持 `database-pools/v1`，只更新 MCP 不能安装后端。
+- `microi_manage_system_observability(action=ResetDatabasePools,poolTarget=Both)` 无确认值时只预览。执行必须带同一预览的 `operationId`、`poolIds`、`poolTarget` 和 `confirmExecution=ResetDatabasePools:<operationId>`。预览票据 2 分钟有效；执行前仍复核配置摘要。
+- 请求超时或响应未知，使用查询动作 `DatabasePoolRecovery` 与原 `operationId` 回读，禁止换编号重试。每租户 60 秒冷却；命令期限 90 秒；回执保留 10 分钟，过期编号不能重新执行。
+- 回执逐节点区分 `Pending / Incomplete / PartialFailure / CompletedForRegisteredNodes`。只有最后一项证明已注册节点完成；未升级节点不在覆盖范围，最后还需验证原故障只读接口。不得把当前节点探测成功说成全站恢复。
+- 支持当前租户 MySQL/SQL Server 主、读池；`Both / Write / Read`，相同池去重。已加载的其它租户共享池时拒绝；扩展库和其它驱动明确不支持。池上限、驱动版本、退避和失败码可读，`Opening` 不等于驱动借出数。
+- 服务继续验证 DiyToken、Redis 有效会话和主库管理员；访问密钥需要 `mcp:admin`。仅鉴权使用有界无池连接，每节点应急请求最多 2 个、连接和命令超时各 5 秒；修复后的探测使用原业务池。不得创建匿名维护引擎或另设万能密码。
+- 不杀事务、不全局清池、不重放 SQL、不自动放大池上限，不以重启数据库/API 作为默认恢复。无有效会话、Redis/主库不可用或内存保护触发时不能绕过防护。持续泄漏、慢 SQL、容量与网络故障仍需定位根因。
+- 此功能属框架运行时应急协议，不新增表/菜单/引擎，不需要发布商城资源；责任同步为 Core/Dos.ORM、MCP、本文档与内嵌知识。验收必须覆盖真实池耗尽、HTTP 撤权/失效会话、两个独立进程、幂等重试、配置漂移与事务存活。
+
+`microi_query_system_observability` 支持：
 
 | action | 用途 | 关键参数 |
 |---|---|---|