How do I install the Obs Review Loop skill?

Run npx skillmds@latest add shenxingy/obs-review-loop in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Obs Review Loop safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Obs Review Loop?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Obs Review Loop free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Obs Review Loop?

shenxingy (@shenxingy) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Obs Review Loop by shenxingy — Agent Skill

Name: Obs Review Loop
Author: shenxingy

Obs Review Loop

Real-user observability review — pull recent Sentry errors, PostHog behavior data, and application logs, exclude internal traffic, triage what real users actually hit, fix obvious bugs, and assess whether the "how do we know this succeeded" monitoring is complete. NOT marketing analytics or a one-off error dump.

shenxingy Updated Sep 21, 2026 8 repo stars

File contents

Observability Review Loop

Turn real-user signals (Sentry, PostHog, app/DB logs) into fixes and into a completeness verdict on the success-mechanism. The executable instructions live in prompt.md; this body is the Codex-facing summary.

Key contracts:

Read-only against observability APIs. Credentials are resolved from the target project's own env/config — never hard-coded, never printed.
Internal traffic (VPN/tailnet IPs, staff accounts, health-check bots, the owner's own devices) is excluded, and the exclusion rule is stated in the report.
Every triaged issue links evidence (issue URL/count, affected flow, likely code location) and lands in one of: 真bug / 体验摩擦 / 数据噪声.
Obvious bugs with clear repro are fixed in a branch, gated, and committed. Muting/ignoring an error without a code fix or a documented rationale is forbidden.
For each fix, "how will we know it worked?" must have an answer; missing metric/alert/dashboard coverage is reported as a 机制缺口, not silently accepted.
Final report is in Chinese and categorized: 已修 (commits + evidence) / 发现未修 (why) / 机制缺口 / 数据健康度.

Frequently asked questions