Investigate First

by juliusbrussee2e08b9177c07No licenseListed Oct 8, 2026Updated Oct 8, 2026

Diagnose ambiguous failures before editing. Use for unknown causes, intermittent behavior, performance regressions, or investigations needing evidence-ranked hypotheses.

Instructions onlyResearch & Analysis
AI-generated overview

Guides an agent to gather evidence and rank hypotheses before editing code when diagnosing ambiguous failures.

What it does
This skill provides an investigation-first procedure for diagnosing failures whose cause is unknown. It directs the agent to separate observed symptoms from inferred causes, trace inputs, state transitions, ownership boundaries, and failure output, and rank hypotheses by evidence and cheap falsification value. It produces a report of the cause and supporting proof, and it explicitly withholds any fix unless the task authorizes implementation.
When to use it
Use it for unknown causes, intermittent behavior, performance regressions, or investigations that need evidence-ranked hypotheses before any code change is made.
Requirements
No tools, packages, runtimes, credentials, or network access are required; the skill is instructions only and ships no scripts.

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.

Source and attribution

Source:juliusbrussee/cavemaninskills/investigate-firstat commit2e08b91

License: No license

Content belongs to its original authors. SourceWeft indexes it from a public repository.

Report or request removal