# TraceCompiler: Skill-Guided Mining and Compilation of LLM Agent Traces into Mostly Deterministic Workflows arXiv: https://arxiv.org/abs/2608.02680 ยท HTML: https://mlsanity.com/paper/2608.02680/ ## Key facts - Cited by 5 papers in the map (data through 14 Oct 2026). - Placed in Software engineering coding agents, one of 16 directions in LLM agents and tool use. - Builds on 19 of its 33 references that are in the map. - First posted on arXiv 3 Aug 2026. ## Placed in - LLM agents and tool use / [Software engineering coding agents](https://mlsanity.com/agents/software-engineering-coding-agents/) ## Builds on (most cited first, 19 of 19) - [AppWorld: A Controllable World of Apps and People for Benchmarking Interactive Coding Agents](https://mlsanity.com/paper/2407.18901/) (2024) - [Agent Workflow Memory](https://mlsanity.com/paper/2409.07429/) (2024) - [An LLM Compiler for Parallel Function Calling](https://mlsanity.com/paper/2312.04511/) (2023) - [Memp: Exploring Agent Procedural Memory](https://mlsanity.com/paper/2508.06433/) (2025) - [ToolMind Technical Report: A Large-Scale, Reasoning-Enhanced Tool-Use Dataset](https://mlsanity.com/paper/2511.15718/) (2025) - [SkillCraft: Can LLM Agents Learn to Use Tools Skillfully?](https://mlsanity.com/paper/2603.00718/) (2026) - [Discovering Object-Centric Petri Nets](https://mlsanity.com/paper/2010.02047/) (2020) - [Parallelizing Tool Execution and LLM Generation for Low-Latency Agent Serving](https://mlsanity.com/paper/2603.18897/) (2026) - [Process Mining Meets Causal Machine Learning: Discovering Causal Rules from Event Logs](https://mlsanity.com/paper/2009.01561/) (2020) - [T1: A Tool-Oriented Conversational Dataset for Multi-Turn Agentic Planning](https://mlsanity.com/paper/2505.16986/) (2025) - [Agentic AI Process Observability: Discovering Behavioral Variability](https://mlsanity.com/paper/2505.20127/) (2025) - [Lifting Traces to Logic: Programmatic Skill Induction with Neuro-Symbolic Learning for Long-Horizon Agentic Tasks](https://mlsanity.com/paper/2605.01293/) (2026) - [Process Discovery Using Graph Neural Networks](https://mlsanity.com/paper/2109.05835/) (2021) - [PreAct: Computer-Using Agents that Get Faster on Repeated Tasks](https://mlsanity.com/paper/2606.17929/) (2026) - [GRADE: Graph Representation of LLM Agent Dependency and Execution](https://mlsanity.com/paper/2606.22741/) (2026) - [SKILL-DISCO: Distilling and Compiling Agent Traces into Reusable Procedural Skills](https://mlsanity.com/paper/2606.26669/) (2026) - [AgentTrails: Towards Trust and Reuse for Agentic Tasks](https://mlsanity.com/paper/2607.18816/) (2026) - [Agentic Compilation: Mitigating the LLM Rerun Crisis for Minimized-Inference-Cost Web Automation](https://mlsanity.com/paper/2604.09718/) (2026) - [WorkflowGen:an adaptive workflow generation mechanism driven by trajectory experience](https://mlsanity.com/paper/2604.19756/) (2026) ## Built on by (most cited first, 5 of 5) - [SkillEffect: Checked Lowering for Memory-Bounded Agent Tools](https://mlsanity.com/paper/2608.17007/) (2026) - [HEXIS: Compiling Agent Skills into Extended Finite State Machines](https://mlsanity.com/paper/2609.30123/) (2026) - [What Can Component-Replacement Evidence Establish? A Critical Scoping Review of Local Decisions in LLM Agents](https://mlsanity.com/paper/2609.39989/) (2026) - [When to Compile a Computer-Use Agent? Measuring Payback and Making Compilation Decisions for Token Efficiency](https://mlsanity.com/paper/2610.02932/) (2026) - [From Traces to Agentic Worlds: Agentic Language World Models for Interactive Environment Simulation](https://mlsanity.com/paper/2610.06100/) (2026) ## Abstract (arXiv) Tool-using language-model agents repeatedly rediscover procedures they have already executed, producing traces that mix reusable structure with retries, exploration, accidental ordering, and repeated lookups. We present TraceCompiler, a skill-guided system that mines clusters of noisy agent traces and compiles them into executable, mostly deterministic workflows. It admits an inter-tool dependency only when a consumer argument contains a value attributable uniquely to an earlier producer; every hard edge carries an auditable evidence tuple, and ambiguous relations are marked suspected and impose no ordering constraint. Bindings are classified as constants, user inputs, copied outputs, transforms, or residual LLM decisions. On T1, a mechanized form of the rule recovers producer-consumer dependencies at 0.928 precision and 0.943 recall over 15,775 def-use edges of its training split, against 0.711 F1 for adjacency and 0.712 for a frequency-thresholded directly-follows measure on identical data; the compiler skill run blind reaches 0.992 on 250 of those edges. On AppWorld we replay released trajectories in the deterministic simulator to recover masked return values and measure the rule against 563 token edges at 0.993 precision - a self-consistency check, since replay injects tokens by a related heuristic. We compile two recurring intents: a Venmo money-request intent reduces 34 observed API calls to 11 runtime calls and, under leave-one-out execution against the benchmark's own state tests, passes 15 of 21, the failing fold escalating rather than acting because its required branch was never observed; and a Spotify/Todoist intent the compiler correctly refuses to compile, because an irreversible side effect is under-determined. We measure call reduction but not offline compilation cost, so we claim no net efficiency result.