一键在 Manus 中运行任何 Skill

investigate

星标33

分支0

更新时间2026年6月8日 06:30

Investigate live or deployed-app issues by gathering evidence from health checks, logs, metrics, deploy history, config, and remote hosts. Use when the problem is happening outside the local dev loop: production, staging, a server, Cloudflare, VPS, CI/CD runtime, logs, alerts, or deployed behavior.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

hunvreus

hunvreus/skill-issue

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

Investigate Live

Input

A live or deployed app, environment, host, service, incident, symptom, alert, or user-reported production issue.
Use explicit input first; otherwise infer from context, logs, alerts, selected files, or branch.
Safest default: identify environment, host/platform, timeframe, and symptom before running commands.

Workflow

Define symptom. Capture what is broken, when it started, who is affected, and what changed recently.
Map targets. Identify platform, environment, host/service names, deploy version, domains, databases, queues, and dependencies.
Gather read-only evidence. Check health endpoints, logs, metrics, recent deploys, config, resource usage, errors, and dependency status.
Form hypotheses. List likely causes and test one at a time.
Use bounded commands. Prefer targeted commands with timeouts and limited output; avoid dumping secrets or huge logs.
Recommend fix path. Separate immediate mitigation, root-cause fix, rollback, and follow-up hardening.
Document findings. Update the relevant docs/development/*.md file if the investigation reveals missing developer or operator knowledge.

Output

Symptom and affected scope
Evidence gathered
Most likely cause
Immediate mitigation
Root-cause fix or next diagnostic step
Docs/runbook gaps

Guardrails

Use debug instead when the issue can be reproduced locally with tests, scripts, or a dev server.
Start read-only; do not restart services, run migrations, change DNS, or mutate production without explicit approval.
Do not expose secrets from logs, env files, dashboards, or command output.
Distinguish facts from hypotheses.
Keep remote targets explicit and commands bounded.

同仓库更多 Skills

同仓库

audit

hunvreus/skill-issue

Audit a new or unfamiliar project to understand its structure, health, risks, documentation gaps, and next actions. Use when initially entering a repo, assessing project quality, preparing onboarding, or asking what should be improved first.

2026-06-0833

debug

hunvreus/skill-issue

Debug failing, flaky, slow, or incorrect behavior with a reproducible feedback loop. Use when something is broken, throwing, failing tests, behaving unexpectedly, or showing a performance regression.

2026-06-0833

deploy

hunvreus/skill-issue

Set up, validate, or run deployment for the current app on Cloudflare, a VPS, or a mixed architecture. Use when the user asks to deploy an app, configure deployment, set up CI/CD, choose a deployment target, document deployment, or fix deployment readiness.

2026-06-0833

document

hunvreus/skill-issue

Create or update project documentation, README sections, architecture docs, API docs, examples, migration notes, devops docs, runbooks, or usage guides. Use when the user asks to document behavior, explain usage, update docs after code changes, or remove stale documentation.

2026-06-0833

handoff

hunvreus/skill-issue

Create a concise handoff for another agent or future session to continue the work without replaying the conversation. Use when the user asks for a handoff, session summary, continuation note, or context transfer.

2026-06-0833

overhaul

hunvreus/skill-issue

Plan and execute a cautious, multi-step improvement effort for a repository or major subsystem. Use when the user asks for a broad cleanup, modernization, technical debt pass, large refactor, or coordinated improvement across architecture, tests, docs, performance, reliability, and maintainability.

2026-06-0833

name	investigate
description	Investigate live or deployed-app issues by gathering evidence from health checks, logs, metrics, deploy history, config, and remote hosts. Use when the problem is happening outside the local dev loop: production, staging, a server, Cloudflare, VPS, CI/CD runtime, logs, alerts, or deployed behavior.

Investigate Live

Input

A live or deployed app, environment, host, service, incident, symptom, alert, or user-reported production issue.
Use explicit input first; otherwise infer from context, logs, alerts, selected files, or branch.
Safest default: identify environment, host/platform, timeframe, and symptom before running commands.

Workflow

Define symptom. Capture what is broken, when it started, who is affected, and what changed recently.
Map targets. Identify platform, environment, host/service names, deploy version, domains, databases, queues, and dependencies.
Gather read-only evidence. Check health endpoints, logs, metrics, recent deploys, config, resource usage, errors, and dependency status.
Form hypotheses. List likely causes and test one at a time.
Use bounded commands. Prefer targeted commands with timeouts and limited output; avoid dumping secrets or huge logs.
Recommend fix path. Separate immediate mitigation, root-cause fix, rollback, and follow-up hardening.
Document findings. Update the relevant docs/development/*.md file if the investigation reveals missing developer or operator knowledge.

Output

Symptom and affected scope
Evidence gathered
Most likely cause
Immediate mitigation
Root-cause fix or next diagnostic step
Docs/runbook gaps

Guardrails

Use debug instead when the issue can be reproduced locally with tests, scripts, or a dev server.
Start read-only; do not restart services, run migrations, change DNS, or mutate production without explicit approval.
Do not expose secrets from logs, env files, dashboards, or command output.
Distinguish facts from hypotheses.
Keep remote targets explicit and commands bounded.