The problem is that agent skills are usually hand-written, made once by an LLM, or revised in loose ways that can easily…
Rohan Paul Twitter · Rohan Paul (@rohanpaul_ai) · 2026-05-29
Microsoft's SkillOpt system proposes training AI agent skills as small optimizable programs rather than hand-writing or one-shot generating them, addressing systematic degradation in current skill-revision workflows.
Appears in
Extraction
Topics: agent-skillsskill-optimizationai-agentsllm-systems
Claims
- Agent skills are typically hand-written or generated once by an LLM without systematic iterative improvement.
- Loosely revising agent skills can easily make them worse rather than better.
- SkillOpt treats agent skills as trainable external programs rather than static prompts or instructions.
Key quotes
The problem is that agent skills are usually hand-written, made once by an LLM, or revised in loose ways that can easily make them worse.
SkillOpt from Microsoft, argues that agent skills should be trained like small external programs, it teaches AI agents better task habits