Diagnose bugs and performance regressions by testing ranked hypotheses against a reproducible feedback loop.
npx skills add https://github.com/kieranpotts/skills --skill diagnose-kieranpotts
Please help me install this Agent Skill. Skill: diagnose Source: https://github.com/kieranpotts/skills/tree/main/skills/diagnose Command: npx skills add https://github.com/kieranpotts/skills --skill diagnose-kieranpotts
High-intent search queries and answers about installing and using this skill.
Build a fast, deterministic feedback loop that reproduces the failure first, then generate 3-5 ranked falsifiable hypotheses and test them one variable at a time. Confirm exactly one cause through direct observation before attempting any fix.
Establish a numerical baseline with a profiler, timing harness, or query plan, then bisect against a threshold. Treat the benchmark as the feedback loop and make the regression test a benchmark assertion rather than a functional test.
No. It stops at a confirmed cause and hands over the evidence: symptom, repro command, ranked hypotheses, causal chain, a failing regression test, and a suggested remedy. Applying the repair is a separate step with its own verification.
Capture a production artifact such as a HAR file, request log, or database snapshot and replay it locally. If no feedback loop can be built at all, the agent stops and reports what it needs rather than guessing at a cause.
Skip it when a compiler, linter, or type-checker has already named the cause, since that is mechanical repair with nothing to diagnose. It is also unnecessary for bugs whose cause is obvious from reading the code.
amrhas82Identify root causes in software bugs using a four-phase investigation framework.
rabbicseConvert vague bug reports into minimal reproductions and failing regression tests.
baasith6Document technical decisions using structured Architecture Decision Records with sequential numbering.
jeremybrasherDiagnose bugs through a four-phase root-cause investigation process before proposing fixes.
