No-ops
person concept tool org talk claim — click a node to jump to its page; hover an arrow for the relation
No-ops (no-operations) are instructions in an agent's skill set that do nothing to change the agent's behavior — they execute without effect, consuming context or processing without contributing to task performance.
The problem with AI-generated skills
Philipp Schmid (Google DeepMind) identifies no-ops as a characteristic failure mode of AI-generated skills. When AI systems automatically generate skills for agents, the resulting instructions frequently include no-ops that pad out the skill definition without altering what the agent actually does. Schmid criticizes this pattern directly: AI-generated skills "tend to include a lot of no-ops," where each is "an instruction which does nothing to change the agent's behavior." Don't Ship Skills Without Evals, 11:14
This issue is part of a broader finding that human-written skills outperform their AI-generated counterparts — Schmid notes that "human-written skills are the best we can provide" and that "AI-generated skills can impact performance negatively." Don't Ship Skills Without Evals, 11:11 No-ops are one concrete mechanism by which AI-generated skills degrade rather than improve agent behavior, making evals essential before shipping any skill to production.