Skip to content

Research

AgentLTL is a language derived from first-order linear temporal logic for expressing procedural rules over agent traces. The same specification can:

  • measure procedural compliance on completed traces, separately from task correctness;
  • enforce it at run time, by gating each tool call before it executes;
  • train for it, as a reward signal for fine-tuning.

AgentLTL: A Trace-Verification Framework for Measuring, Enforcing, and Training Procedural Compliance in Tool-Using LLM Agents
Laïla Elkoussy, Julien Perez. arXiv:2607.02599, 2026.

@misc{elkoussy2026agentltl,
  title         = {AgentLTL: A Trace-Verification Framework for Measuring, Enforcing, and Training Procedural Compliance in Tool-Using LLM Agents},
  author        = {Elkoussy, La{\"\i}la and Perez, Julien},
  year          = {2026},
  eprint        = {2607.02599},
  archivePrefix = {arXiv},
  primaryClass  = {cs.SE},
  url           = {https://arxiv.org/abs/2607.02599}
}

Repositories

Repository What it is
AgentLTL The formula engine, post-hoc verification, and run-time enforcement for smolagents or any agent loop
cli-to-tools Shell command lines → ordered, structured tool calls
agentltl-coding The rule language, rule library and guard that every coding-agent harness shares
agentltl-claude-code The Claude Code harness
agentltl-copilot-cli The GitHub Copilot CLI harness
agentltl-mistral-vibe The Mistral Vibe harness
agentltl-codex The Codex harness
AgentLTL.github.io This site

Contact

Laïla Elkoussy · laila.elkoussy@epita.fr