dsh-observability：治理套件共用诊断日志与运行诊断页
dsh-observability 为 DeepSeek Harness Claw 治理插件提供结构化诊断与 Web「设置→运行诊断」页。其它插件经 observe.mjs 只记录操作终态（succeeded/rejected/failed/degraded），字段白名单排除 prompt 与 token；日志写入 $DSH_HOME/agent-observe/diag.jsonl，安全审计仍由 dsh-agent-gate 负责。安装：dsh plugin --profile web add github:xingyingyuzhui/dsh-observability，重启 dsh web 后使用。
Read Moredsh-harness-trace：把 Harness 与 LLM 的整段交互落成可读时间线
oihai 维护的 DSH 插件 dsh-harness-trace 只读监听 Harness 事件，记录完整提示词、技能、工具、文件、联网与 Token 等，按轮次与步骤汇总为高可读 Markdown，默认写入会话工作区 harness-trace 目录。提供侧边栏审阅页、设置与会话内入口，可配置采集项、详细程度与追踪范围。数据不依赖其他插件；README 说明以 link 指向源码开发安装，且仅在 Desktop 完成测试。
Read Moredsh-health-readout：用声明式配置把多处读数合成一条健康结论
dsh-health-readout（laa1991/dsh-health）是 DeepSeek Harness 上的只读工作流插件：在 readings.json 中声明数据源、healthy 条件与 blind 说明，通过工具 health_readout 组合文件、JSON、JSONL 尾部与 HTTP JSON 等读数，按 ok→broken 严重度输出一条总判决。支持 selftest 验证 criterion 可证伪；配置 dataDir/specFile 独立于 DSH 平台目录。MIT，0.1.0，Node≥18。
Read MoreObservabilityAndInstrumentation：让 Agent 按流程完成可观测性与埋点
ObservabilityAndInstrumentation 是 SkillHub 社区 Skill（@user_3c6cb52e/observability-and-instrumentation-sh-k1xmdy，v1.0.0，it-ops-security，无需 API Key）。包内 SKILL.md 指导 Agent 在上线前定义值班问题，按 RED/USE 落地结构化日志、指标、OpenTelemetry 链路与症状型告警，并用验收清单验证遥测。可通过 skillhub CLI 安装到 Cursor、Claude Code、Codex 等 skills 目录。本文基于官方页面与注册表下载包核对功能与安装命令。
Read MoreDeepSeek Harness Plugin: CacheScope - Prompt Caching Observability and Input Diagnostics
CacheScope is a local debugging plugin for DeepSeek Harness (DSH) designed to address the challenge of locating low prompt cache hit rates. It provides a desktop dashboard that displays Provider-reported cache token usage in real time, along with diagnostic features such as comparing two dialogue turns, inspecting differences between adjacent inputs, and visualizing a color-coded input tree. Data can be filtered by session, model, and other criteria. Installation is straightforward, and by default, the plugin retains only the most recent 12 inputs. Note: Data exists only in local memory and cannot be used for cross-session comparisons after restarting the process.
Read Moredsh-maestro-observe: Local-First Observability for the Maestro Plugin Stack
`dsh-maestro-observe` is a local, lightweight observability tool designed specifically for the DeepSeek Harness and Maestro plugin stacks. It uses SQLite for local storage of traces, costs, latency, and health status, and provides CLI and RPC endpoints for querying. The plugin supports analysis grouped by tool or session, and includes features such as budget control and error monitoring. Developers should be aware of its Node.js version dependency and the fact that sensitive information is redacted before data is persisted.
Read Moresatan9394/dsh-observability-tools: Observability Tools
dsh-observability-tools is an observability skill plugin designed for DSH environments and maintained by satan9394. It integrates Prometheus, Grafana, and tracing capabilities to address the needs for monitoring and call-chain tracking in agent development. Developers can find installation guidance on its catalog page. When using it, pay attention to permissions to ensure the plugin runs with the current process's permissions, and review the source code for compliance when license information is unclear.
Read Moredsh-observability: DeepSeek Harness Observability Skill Plugin
`satan9394/dsh-observability` is an observability plugin for DeepSeek Harness (DSH), intended to help developers monitor agent workflows through logs, metrics, and traces (the three pillars). Inspired by agent-skills, the plugin currently does not provide specific features, installation commands, or usage examples. Because it runs with process-level permissions, developers should carefully review the source code for security and compliance.
Read MoreDeepSeek Harness Litefuse Plugin: Agent Observability and Evaluation
The Litefuse plugin is the official observability plugin for DeepSeek Harness, designed to address tracing challenges when debugging complex workflows and subagent invocations. It transforms DSH runtime data into Litefuse-compliant traces, supporting turn-level mapping, subagents, code modes, and zero-dependency OTLP integration. By installing it via `npx` and configuring the public and private keys in environment variables, users can view real-time execution details on the Litefuse platform, including latency, token usage, and context compression events. The plugin runs with the permissions of the current process, reads environment variables by default, and strictly safeguards key security.
Read Moredsh-observability: Add attribution for DSH external sub-agent calls
dsh-observability is an MIT-licensed plugin maintained by jeremy9682, aimed at solving the observability issue of verifying session outputs after DSH invokes external sub-agents (such as Codex and Cursor). The repository contains two core plugins: dsh-codex-observability and dsh-tool-cursor. The former tracks directories such as `~/.codex/sessions` via fingerprinting, writes attribution pointers like `sessionRef` into journal files, recording only the pointers without storing prompt or transcript content. The latter provides the `cursor_delegate` tool, supporting the delegation of tasks to the Cursor CLI, and defaults to read-only but supports editing.
Read Moredsh-o11y-plugin: Unified plugin-level observability for deepseek-harness
dsh-o11y-plugin is an observability plugin designed specifically for deepseek-harness (dsh), aiming to unify telemetry management at the plugin level. By registering global OpenTelemetry providers, it enables plugins based on the standard @opentelemetry/api to conveniently export traces, metrics, and logs, while bridging DSH session telemetry into OTel logs. The plugin supports both headless and web profiles, exporting data via OTLP/HTTP endpoints; if no endpoint is configured or it is unreachable, data will be silently discarded. Stateless by design, the plugin requires no data cleanup upon removal, does not write to files or store credentials, and ensures telemetry data resides in memory only before export.
Read Moredsh-plugin-langfuse: Exporting DSH Agent Sessions to Langfuse
dsh-plugin-langfuse is a community plugin for DeepSeek Harness maintained by linyp, implementing session telemetry integration. It exports each round of conversation as OpenTelemetry traces compliant with GenAI semantics, writes them into Langfuse via OTLP/HTTP, and supports feedback Scores, session aggregation, and fork lineage. This article introduces its positioning, core capabilities, profile bundle installation steps, environment variables and mode configuration, as well as considerations such as regional keys, privacy controls, and runtime permissions.
Read More# Sentry Skill: Integrating Production Error Troubleshooting into AI Agent Workflows
Sentry is an Agent Skill maintained by OpenAI in the curated directory of the openai/skills repository, which enables read-only observability queries via the official Sentry CLI: listing unresolved issues, viewing event details, and calling issue explain/plan to generate root cause analysis and repair suggestions. This article introduces its positioning, core commands, installation methods for both the CLI and the Skill (Codex's $skill-installer and Cursor's .cursor/skills), as well as precautions including the default 24-hour production environment query scope, JSON output format and security desensitization, which is suitable for DevOps/SRE to integrate online troubleshooting into AI workflows.
Read More