Search papers, labs, and topics across Lattice.
This study investigates the conditions under which skills enhance the performance of LLM agents, identifying key factors such as representation, outcome annotation, and retrieval difficulty. Through a combination of controlled experiments and trajectory analysis, the authors establish a taxonomy of skill-use modes, revealing that procedural anchoring is the primary mechanism through which skills stabilize agent execution. Notably, skills improve performance metrics by 6.06 points, while also highlighting the challenges posed by retrieval bottlenecks and the limitations of skill application in certain contexts.
Skills stabilize agent execution by transforming noisy trajectories into procedural anchors, but they can fail under brittle assumptions and incompatible contexts.
Skills have emerged as a practical and effective approach for enhancing LLM agents at inference time through structured packages of knowledge. However, existing evaluations largely measure whether skills improve aggregated task success, leaving a more fundamental question underexplored: \emph{\textbf{When do skills help, why do they work, and where do they fail?}} Through controlled experiments across various benchmarks, agent harnesses and LLMs, we isolate the effects of representation, outcome annotation, retrieval difficulty, and cross-framework robustness of skills. To further answer this question, we design a contrastive study that combines controlled quantitative experiments with paired trajectory analysis. We normalize 8,135 trial records from controlled experiments and retain 238 valid unique labels from 240 open-coded records. We consolidate these observations into a taxonomy of three high-level categories and twelve skill-use modes: skills work when noisy trajectories become procedural anchors that stabilize execution. Skills improve over Workflow Memory by 6.06 points in matched comparisons. Procedural anchoring accounts for 65.7\% of skill cases, versus 4.5\% for explicit knowledge injection, showing that skills stabilize action rather than inject missing facts. Retrieval is a separate bottleneck: as pools grow from 5 to 100, actual-use precision falls from 29.6\% to 3.3\%. Confusable distractors impair offline identification, yet downstream success remains stable; exact ground-truth invocation is neither sufficient nor necessary. Skills fail under brittle assumptions, incompatible contexts, or insufficient adaptation. These findings move evaluation beyond aggregate success rates and guide reliable self-evolving agents.
