sf-ai-agentforce-observability by jaganpro/sf-skills
npx skills add https://github.com/jaganpro/sf-skills --skill sf-ai-agentforce-observability当用户需要基于追踪的可观测性而不仅仅是测试时,请使用此技能:提取会话追踪数据模型记录、处理 Parquet 数据集、重建会话时间线、分析主题/操作延迟,或基于 Data 360 遥测数据调试智能体行为。
当工作涉及以下内容时,使用 sf-ai-agentforce-observability:
.parquet 文件当用户进行以下操作时,请委托给其他技能:
提取前,请验证:
广告位招租
在这里展示您的产品或服务
触达数万 AI 开发者,精准高效
如果缺少身份验证,请转交给:
深度设置指南:
至少,预期工作围绕以下内容展开:
GenAI 信任层/审计记录也可能与内容质量和生成调试相关。
完整模式:
询问或推断:
确认 Data 360 追踪存在且 JWT/ECA 身份验证正常工作。
| 需求 | 默认方法 |
|---|---|
| 最近的遥测快照 | 提取最近 N 天 |
| 针对性调查 | 按日期和智能体进行筛选提取 |
| 单个故障对话 | 提取或调试单个会话树 |
| 持续使用情况分析 | 增量提取 |
使用 scripts/ 目录下提供的脚本,而不是重新实现提取逻辑。
常见分析目标:
典型结果:
常见陷阱:
完成时,按此顺序报告:
建议格式:
Observability task: <extract / analyze / debug-session>
Scope: <org, dates, agents, session ids>
Artifacts: <directories / parquet files>
Findings: <latency, routing, action, quality, abandonment patterns>
Root cause: <best current explanation>
Next step: <testing, agent fix, flow fix, apex fix>
| 需求 | 委托给 | 原因 |
|---|---|---|
| 身份验证 / JWT 设置 | sf-connected-apps | Data 360 访问 |
| 修复智能体路由/行为 | sf-ai-agentscript | 编写修正 |
| 正式的回归/覆盖率测试 | sf-ai-agentforce-testing | 可重复的测试循环 |
| Flow 支持的操作调试 | sf-flow | 声明式修复 |
| Apex 支持的操作调试 | sf-debug 或 sf-apex | 代码/日志调查 |
| 分数 | 含义 |
|---|---|
| 90+ | 基于遥测数据的强力诊断 |
| 75–89 | 有用的分析,存在微小差距 |
| 60–74 | 仅部分可见性 |
| < 60 | 证据不足;收集更多遥测数据 |
每周安装次数
230
代码仓库
GitHub 星标数
223
首次出现
Jan 29, 2026
安全审计
安装于
codex226
opencode226
cursor226
gemini-cli224
github-copilot224
amp223
Use this skill when the user needs trace-based observability , not just testing: extract Session Tracing Data Model (STDM) records, work with Parquet datasets, reconstruct session timelines, analyze topic/action latency, or debug agent behavior from Data 360 telemetry.
Use sf-ai-agentforce-observability when the work involves:
.parquet files from Agentforce telemetryDelegate elsewhere when the user is:
Before extraction, verify:
If auth is missing, hand off to:
Deep setup guide:
At minimum, expect work around:
GenAI Trust Layer / audit records may also be relevant for content-quality and generation debugging.
Full schema:
Ask for or infer:
Confirm Data 360 tracing exists and JWT/ECA auth is working.
| Need | Default approach |
|---|---|
| recent telemetry snapshot | extract last N days |
| focused investigation | filtered extraction by date and agent |
| one broken conversation | extract or debug a single session tree |
| ongoing usage analytics | incremental extraction |
Use the provided scripts under scripts/ rather than reimplementing extraction logic.
Common analysis goals:
Typical outcomes:
Common pitfalls:
When finishing, report in this order:
Suggested shape:
Observability task: <extract / analyze / debug-session>
Scope: <org, dates, agents, session ids>
Artifacts: <directories / parquet files>
Findings: <latency, routing, action, quality, abandonment patterns>
Root cause: <best current explanation>
Next step: <testing, agent fix, flow fix, apex fix>
| Need | Delegate to | Reason |
|---|---|---|
| auth / JWT setup | sf-connected-apps | Data 360 access |
| fix agent routing / behavior | sf-ai-agentscript | authoring corrections |
| formal regression / coverage tests | sf-ai-agentforce-testing | reproducible test loops |
| Flow-backed action debugging | sf-flow | declarative repair |
| Apex-backed action debugging | sf-debug or sf-apex |
| Score | Meaning |
|---|---|
| 90+ | strong telemetry-backed diagnosis |
| 75–89 | useful analysis with minor gaps |
| 60–74 | partial visibility only |
| < 60 | insufficient evidence; gather more telemetry |
Weekly Installs
230
Repository
GitHub Stars
223
First Seen
Jan 29, 2026
Security Audits
Gen Agent Trust HubPassSocketPassSnykPass
Installed on
codex226
opencode226
cursor226
gemini-cli224
github-copilot224
amp223
AI Elements:基于shadcn/ui的AI原生应用组件库,快速构建对话界面
58,500 周安装
Next.js 最佳实践指南:App Router 开发原则、性能优化与项目结构
3,800 周安装
Reka UI - 无样式、可访问的 Vue 3 组件库 | 符合 WAI-ARIA 标准
3,800 周安装
核心Web指标优化指南:LCP、INP、CLS三大性能指标详解与实战技巧
3,900 周安装
Angular 路由配置指南:v20+ 懒加载、Signal 参数与函数式守卫
3,800 周安装
Angular Signal Forms 教程:构建类型安全响应式表单的完整指南
3,900 周安装
React现代化升级指南:版本迁移、Hooks转换、并发特性与自动化重构
4,000 周安装
| code / log investigation |