Claude MarketplacesSubmit
juliusbrussee avatar

Investigate First

juliusbrussee/caveman

Investigate-first is a guardrail for the moment when Claude Code wants to start editing before it knows what's actually broken. It pushes the agent to separate the observed symptom from the inferred cause, trace inputs, state transitions, and ownership boundaries, and rank candidate hypotheses by how much evidence supports them and how cheaply each can be falsified. No product code changes until one mechanism credibly explains the evidence, and the report is expected to name the cause and the proof rather than a speculative patch. I'd reach for it on intermittent failures, performance regressions, and anything where the cause is unknown, since those are where an eager agent does the most damage by "fixing" the wrong thing. The honest caveat is that it's a short prompt-level skill, so its discipline depends on the agent following it, and it works best when you explicitly say whether you want a diagnosis only or a fix.

skills.sh installs
120.1K
GitHub stars
118.7K
last updated
Oct 26, 2026
View on GitHub

Install

$ npx -y skills add juliusbrussee/caveman --skill investigate-first --agent claude-code

Installs into .claude/skills of the current project.

Files

Investigate first

Gather evidence before changing product code.

  • Separate observed symptom from inferred cause.
  • Trace inputs, state transitions, ownership boundaries, and failure output.
  • Rank hypotheses by evidence and cheap falsification value.
  • Do not edit until one credible mechanism explains evidence.
  • Stop exploration when evidence is sufficient to name cause or exact blocker.

Report cause and proof. Make no fix unless task authorizes implementation.

SKILL.md on GitHub

Sponsors

Promote your product hereReach 125k+ visitors monthlyPromote your product hereReach 125k+ visitors monthlyAgent CaffeineSponsoredClose your lid, come back to finished workSnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthlyWaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthlyknobsSponsoredSee your site on any phone, any day, in any language

Details

Repository
github.com/juliusbrussee/caveman
Category
Debugging
First seen
Oct 26, 2026

More from juliusbrussee/caveman

All 25

Recommended

More Debugging
1currents-dev avatarplaywright-best-practicescurrents-dev/playwright-best-practices-skillUse when writing Playwright tests, fixing flaky tests, debugging failures, implementing Page Object Model, configuring CI/CD, optimizing performance, mocking AP104.5Kskills.sh installs
2microsoft avatarazure-diagnosticsmicrosoft/github-copilot-for-azureDebug Azure production issues on Azure using AppLens, Azure Monitor, resource health, and safe triage. WHEN: debug production issues, troubleshoot app service, 103.1Kskills.sh installs
3mcollina avatarfastify-best-practicesmcollina/skillsGuides development of Fastify Node.js backend servers and REST APIs using TypeScript or JavaScript. Use when building, configuring, or debugging a Fastify appli94.4Kskills.sh installs
4microsoft avatarazure-messagingmicrosoft/github-copilot-for-azureTroubleshoot and resolve issues with Azure Messaging SDKs for Event Hubs and Service Bus. Covers connection failures, authentication errors, message processing 92.4Kskills.sh installs
5squirrelscan avataraudit-websitesquirrelscan/skillsAudit a website with the squirrelscan CLI and fix the findings in code. Runs SEO, performance, security, technical, content, accessibility, and 15 other rule ca74.9Kskills.sh installs
6addyosmani avatarbrowser-testing-with-devtoolsaddyosmani/agent-skillsTests in real browsers via Chrome DevTools MCP. Use when building or debugging anything that runs in a browser. Use when you need to inspect the DOM, capture co67.7Kskills.sh installs