@mattshumer_: For those who want to keep their skills while using Opus 5, here's a trick you can try (let me know how it goes): Write…

X AI KOLs Timeline News
claude-opus-5 ai-workflow skills benchmarking tips

Summary

Matt Shumer shares a technique for adapting skills to Claude Opus 5: a loop that updates skills, benchmarks them against Opus 4.8 with old skills, and iterates until Opus 5 wins.

For those who want to keep their skills while using Opus 5, here's a trick you can try (let me know how it goes): Write a loop (have a model do this!) that has Opus 5: - do a first update pass on your skills to make them better for Opus 5 - creates a suite of test tasks for each skill, like a benchmark - runs each test task on Opus 5 with the new skill, and Opus 4.8 with the old one - has a separate model blind-compare the outcomes (Gauntlet Loop-style) - repeatedly iterate until Opus 5 + updated skills is consistently better on the suites than Opus 4.8 + old skills
Original Article
View Cached Full Text

Cached at: 08/08/26, 11:10 PM

For those who want to keep their skills while using Opus 5, here’s a trick you can try (let me know how it goes):

Write a loop (have a model do this!) that has Opus 5:

  • do a first update pass on your skills to make them better for Opus 5
  • creates a suite of test tasks for each skill, like a benchmark
  • runs each test task on Opus 5 with the new skill, and Opus 4.8 with the old one
  • has a separate model blind-compare the outcomes (Gauntlet Loop-style)
  • repeatedly iterate until Opus 5 + updated skills is consistently better on the suites than Opus 4.8 + old skills

Matt Shumer (@mattshumer_): Claude Opus 5 is way better than people give it credit for.

If you’re using it like previous Claude models, it’s going to suck.

Two things make a huge difference:

  • Delete ALL your skills/MCPs/Claude.md/etc. Start fresh.
  • Stop telling it how to do the thing. Just say what you

Similar Articles