skills.rest
ThorStarlord
ThorStarlordCOMMUNITY
Explore all skills by this author →
GitHub

diagnose

Convert vague bug reports into reproducible failure signals and evidence-driven fixes.

1|Updated May 6, 2026
Search Tags:#debugging#instrumentation#regression testing#bisect#performance regression#falsifiable hypotheses#diagnosis loop
One-click install
npx skills add https://github.com/ThorStarlord/interface-skills --skill diagnose-thorstarlord
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: diagnose
Source: https://github.com/ThorStarlord/interface-skills/tree/main/skills/diagnose
Command: npx skills add https://github.com/ThorStarlord/interface-skills --skill diagnose-thorstarlord

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Diagnose fixes hard bugs and performance regressions by forcing a disciplined, evidence-driven workflow instead of guesswork.

Core Features & Use Cases

  • Feedback-loop first: Build a deterministic pass/fail signal (tests, API/CLI scripts, headless browser assertions, trace replay, or a minimal harness) to make the bug debuggable.
  • Reproduce before changing: Confirm the failure mode matches the user report and can be repeated reliably (or at a high failure rate for flaky issues).
  • Ranked, falsifiable hypotheses: Generate 3–5 predictions and only then test them with targeted instrumentation.
  • Instrument, then fix with regression: Change one variable at a time, validate with a minimized repro, and lock in correctness with a regression test when a proper seam exists.
  • Cleanup and post-mortem: Remove instrumentation, validate the original scenario, and document the root cause and prevention strategy (including handoff when architecture blocks test seams).

Quick Start

Ask the agent to diagnose this by using a pass/fail feedback loop, reproducing the reported failure mode, ranking falsifiable hypotheses, instrumenting with targeted probes, applying the fix with a regression test if possible, and confirming the original scenario no longer reproduces.

Frequently Asked Questions about diagnose

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I debug a hard bug that only crashes intermittently?▼

Debugging intermittent crashes requires building a deterministic pass/fail feedback loop, such as a test harness or trace replay, to reproduce the failure mode at a high rate. Generate ranked falsifiable hypotheses, then use targeted instrumentation to isolate the root cause.

What is the best way to find a performance regression in my code?▼

Finding a performance regression involves creating a reproducible signal using API scripts or headless browser assertions. Establish a baseline, generate falsifiable hypotheses, change one variable at a time with controlled instrumentation, and validate the fix with regression coverage.

How do I reproduce a vague bug report before attempting a fix?▼

Reproducing a vague bug report requires selecting an appropriate feedback loop like tests, replay scripts, or headless automation to reach a stable debugging signal. Confirm the failure mode matches the user report and can be repeated reliably before making any code changes.

What should I do when my architecture blocks writing regression tests for a bug fix?▼

When architecture blocks test seams, apply targeted instrumentation to isolate the issue, apply the fix, and complete the workflow with post-mortem guidance. Document the root cause and prevention strategy, including handoff notes for architectural improvements needed to enable future regression tests.

How does bisecting help with diagnosing software regressions?▼

Diagnosing software regressions with a bisect approach uses an evidence-driven workflow to convert vague symptoms into a testable failure signal. By changing one variable at a time and validating against a minimized repro, you rank falsifiable hypotheses to pinpoint the exact regression.

Can I use headless browser automation to debug failing behavior?▼

Yes, headless browser assertions provide a deterministic pass/fail signal to debug failing behavior. This feedback loop helps reproduce the exact failure mode, allowing you to instrument with targeted probes, apply the fix with regression coverage, and confirm the scenario no longer reproduces.

Related Skills

Explore More Software Engineering Skills→
Ilenburg1993Ilenburg1993
3

reactive-bug-audit

Analyze stack traces and logs to identify root causes and apply minimal patches.

Community
Intermediate
NachoOsellaNachoOsella
6

bug-fixer

Identify, reproduce, isolate, fix, verify, and document software bugs.

Community
Advanced
janpeterdjanpeterd

diagnose

Diagnose bugs and performance regressions using deterministic pass/fail feedback loops.

Community
Advanced
nseluganseluga

diagnosing-bugs

Diagnose bugs and performance regressions using structured feedback loops and minimal repro scenarios.

Community
Advanced