When Experience Becomes Instruction: Trajectory Poisoning in Self-Evolving Agent Skill Systems
Self-evolving skill (SES) systems distill agent trajectories into persistent skills, allowing untrusted experience to become trusted instruction. We introduce PoisonedEvolution, a trajectory-poisoning attack on this promotion process. Our skill-visible black-box attacker can inspect a target skill and contribute bounded evidence, but cannot observe private pools or evolution logic or edit the skill bank. Artifact poisoning requires Inclusion, Evolution Attribution, and Realization. Attribution i
Record details
Published: 6 August 2026
Source: arXiv
Category: Research
Topics: Agents & autonomy
Retrieved: 7 August 2026
Related evidence
These records share source-supplied organisations, an exact publisher byline, automatic topics or regions. The reason is shown on every link; related does not mean supporting, agreeing with or verifying this record.
Learning Sexism Detection Using Multi-Agent Perspectivist Preference Optimization
arXiv cs.CY · 6 August 2026
Artificial Institutions: How Institutional Design Shapes LLM Simulations
arXiv cs.CY · 6 August 2026
Hidden Underbelly of the Silicon Valley: Algorithmic Exploitation and Health in Data Work Value Chains
arXiv cs.CY · 6 August 2026
Governing Execution Risk in Agentic AI Systems: A Trajectory-Guided Framework for Red Teaming
arXiv cs.CY · 6 August 2026
PromptShield Home: Ambient Multimodal Prompt Injection Defense for Smart-Home Agents
arXiv cs.HC · 6 August 2026
Contact-aware multi-skill learning framework with hybrid force-motion control for stability and force regulation in robotic physiotherapy
Frontiers in Robotics and AI · 6 August 2026
How to cite this record
ethics.ai (6 August 2026), “When Experience Becomes Instruction: Trajectory Poisoning in Self-Evolving Agent Skill Systems,” evidence record 17092, https://ethics.ai/record/17092 (originally published by arXiv).
Use and limitations
This page is a stable index and citation surface for a source record. ethics.ai did not author the underlying report and has not independently verified every claim. Automatic topics may be imperfect. For consequential use, quote and cite the original publisher.