Agent skills by task
Agent Skills for Debugging Hard Bugs
Debugging skills stop an agent from guessing. They make it reproduce the failure, narrow it down and test one hypothesis at a time, so fixes land on the real cause instead of the nearest symptom.
- skills
- 5
- packs
- 2
The skills
In the order we would try them
-
/diagnosing-bugsA gated debugging loop for hard bugsWorks through hard bugs and slowdowns in gated steps, starting from a reproducible failure.
-
/systematic-debuggingFinds the root cause of a bug in four phases before attempting a fix.
-
/verification-before-completionRequires running checks and reading their output before claiming work is done.
-
/diagnosing-superpowersAnalyzes a session transcript to explain what went wrong, with optional bug report.
-
/wait-whatMakes the agent re-explain its last message in plain language with missing context.
Questions
What makes a debugging skill different from just asking the agent to fix a bug?
The skill fixes the order of work: reproduce, minimise, form hypotheses, test them, then fix with a regression test. Without it, agents tend to patch the first plausible line.
Do these skills work on performance problems?
/diagnosing-bugs covers slowdowns as well as failures: it asks for a baseline measurement before changing code.
Skills for other jobs
Lists picked by trendhub from public skill packs. Each skill lives in its author’s repository; read the current text there before installing.