Microsoft Paper Reveals Hidden Costs of Bad Skills in AI Agent Harnesses
omarsar0 · x · 2026-08-13
Major agent harnesses rely on skill libraries for extra guidance, assuming it's free. A new paper from Microsoft and colleagues measures what a bad skill actually costs.
Researchers attributed 307 agent failures to specific loaded skills (125 functional failures and 182 efficiency regressions) by comparing each skill-guided run against a matched reference run.
- Root of failures: Failures rarely come from irrelevant skills. Seemingly relevant skills often push the agent to incorrectly implement or omit task requirements.
- Efficiency costs: Cost regressions aren't explained by prompt length alone. The largest source is excessive verification (67 cases), followed by heavy implementation pipelines (30 cases). Skills quietly turn validation checklists into mandatory work.
More from coding & agent
- Devs say LLM coding mistakes are a different class: overcomplication, inferred limits, worse at higher reasoning — Mu5_ · 2026-10-16
- How do you debug bad agent runs when you already have tracing? — Sensitive-Parsnip-12 · 2026-10-16
- Where does pre-task context live for agents? Devs debate workflows beyond handoff — shishen2026 · 2026-10-16
- No-instruction autonomous agent wakes up, sets own goals, and tries hacking by default — sethlazar · 2026-10-16
- CODESTRUCT reframes codebases as AST action spaces to fix brittle LLM code agents — burkov · 2026-10-16
- Would you install another SDK for agent debugging on top of your tracing stack? — Sensitive-Parsnip-12 · 2026-10-16