Search papers, labs, and topics across Lattice.
This paper introduces SIGIL, a method for compiling prose-based agent skills into executable harnesses, addressing the inefficiencies of AI-integrated agents that often fail to execute mandated steps. By utilizing a typed agentic intermediate representation (AG-IR), SIGIL enhances the performance of compiled harnesses, achieving an 86% completion rate of mandated steps compared to only 56% for prose skills. The results demonstrate that compiled harnesses not only improve execution consistency across model generations but also reduce token usage by nearly half, highlighting a significant advancement in agent skill utilization.
Compiled harnesses in SIGIL boost execution of mandated steps to 86%, outperforming prose skills by a staggering 30%.
AI-Integrated agents increasingly acquire capability from skills: prose procedure files loaded into a model's context and run by a tool-calling loop. A skill is described to the runtime but never encoded in it, so the model re-derives its control flow on every run and may skip mandated verification. Across 30 skills and two model generations, a prose agent performs only 56% of the steps its own skill mandates, while producing artifacts that pass output checks. The remedy is known: write a harness, in which the procedure is program structure. However, hand-writing harnesses is tedious and discards the authoring surface that made skills succeed. To address this limitation, we introduce Skill Compilation, realized in SIGIL, which compiles a prose skill into an executable harness. At its center is AG-IR, a typed agentic intermediate representation separating model-owned cognition from code-owned mechanism. Compiled harnesses perform 86% of mandated steps, complete the full procedure 2.3x as often, and require 0.58x the tokens. Notably, the guarantee is model-independent: the harness holds at 86% across two model generations while prose swings from 56% to 68%.
