Can this skill fix production issues automatically?

While it can propose and apply fixes, the protocol emphasizes 'Demand Observable Proof,' requiring specific log entries or metric changes to verify a resolution before it is considered resolved.

How does the Investigate skill handle large log files?

It employs a bounded shell output strategy, starting with small slices and counts (e.g., tail -n 200) and utilizing safe-read scripts to prevent terminal flooding or token exhaustion.

Does it integrate with my existing observability stack?

Yes, it is designed to utilize standard tools like sentry-cli, vercel, and convex, as well as local git history and environment configuration lists to gather evidence.

What is Multi-Hypothesis Mode?

When an issue has multiple plausible causes, this mode assigns different AI agents to investigate each hypothesis simultaneously, preventing the system from anchoring on a single incorrect theory.

SRE Incident Investigator

Name: SRE Incident Investigator
Author: phrazzld

byphrazzld

•

分析と監視

Streamlines production incident response through automated root cause analysis, log investigation, and evidence-driven debugging.

The Investigate skill transforms Claude into a senior Site Reliability Engineer capable of diagnosing complex production issues through a systematic, evidence-based protocol. It prioritizes configuration checks over code changes, utilizes multi-agent teams for competing hypotheses, and maintains a real-time incident log (INCIDENT-timestamp.md). By enforcing bounded shell outputs and demanding observable proof for every fix, it ensures that production environments are stabilized safely and documented thoroughly, culminating in a visual incident timeline for postmortem analysis.

主な機能

01Automated incident logging and real-time documentation generation

02Config-first diagnostic protocol to rule out environment issues early

03Visual HTML timeline generation for comprehensive incident postmortems

04Multi-hypothesis mode for diagnosing complex, multi-service failures

05Safe log exploration with bounded shell outputs and read-only operations

065 GitHub stars

ユースケース

01Performing root cause analysis after a failed deployment or configuration change

02Coordinating multi-service investigations using AI agent teams to avoid cognitive bias

03Debugging intermittent production errors or failing webhooks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add phrazzld/claude-config investigate

For use in Claude.ai and ChatGPT

Download Skill