# When Experience Becomes Instruction: Trajectory Poisoning in Self-Evolving Agent Skill Systems

arXiv: https://arxiv.org/abs/2608.05563 · HTML: https://mlsanity.com/paper/2608.05563/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use.
- Builds on 18 of its 24 references that are in the map.
- First posted on arXiv 6 Aug 2026.

## Placed in

- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)
- Synthetic data and self-improvement for LLMs / [Safety of self-evolving systems](https://mlsanity.com/synth/safety-of-self-evolving-systems/)

## Builds on (most cited first, 18 of 18)

- [Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection](https://mlsanity.com/paper/2302.12173/) (2023)
- [The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions](https://mlsanity.com/paper/2404.13208/) (2024)
- [Defeating Prompt Injections by Design](https://mlsanity.com/paper/2503.18813/) (2025)
- [SecAlign: Defending Against Prompt Injection with Preference Optimization](https://mlsanity.com/paper/2410.05451/) (2024)
- [Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills](https://mlsanity.com/paper/2603.25158/) (2026)
- [MemSkill: Learning and Evolving Memory Skills for Self-Evolving Agents](https://mlsanity.com/paper/2602.02474/) (2026)
- [CoEvoSkills: Self-Evolving Agent Skills via Co-Evolutionary Verification](https://mlsanity.com/paper/2604.01687/) (2026)
- [AutoSkill: Experience-Driven Lifelong Learning via Skill Self-Evolution](https://mlsanity.com/paper/2603.01145/) (2026)
- [SpreadsheetBench: Towards Challenging Real World Spreadsheet Manipulation](https://mlsanity.com/paper/2406.14991/) (2024)
- [Skill-Inject: Measuring Agent Vulnerability to Skill File Attacks](https://mlsanity.com/paper/2602.20156/) (2026)
- [SkillClaw: Let Skills Evolve Collectively with Agentic Evolver](https://mlsanity.com/paper/2604.08377/) (2026)
- [Supply-Chain Poisoning Attacks Against LLM Coding Agent Skill Ecosystems](https://mlsanity.com/paper/2604.03081/) (2026)
- [BadSkill: Backdoor Attacks on Agent Skills via Model-in-Skill Poisoning](https://mlsanity.com/paper/2604.09378/) (2026)
- [MUSE-Autoskill: Self-Evolving Agents via Skill Creation, Memory, Management, and Evaluation](https://mlsanity.com/paper/2605.27366/) (2026)
- [Poison Once, Exploit Forever: Environment-Injected Memory Poisoning Attacks on Web Agents](https://mlsanity.com/paper/2604.02623/) (2026)
- [Model Context Protocol Threat Modeling and Analyzing Vulnerabilities to Prompt Injection with Tool Poisoning](https://mlsanity.com/paper/2603.22489/) (2026)
- [RouteGuard: Internal-Signal Detection of Skill Poisoning in LLM Agents](https://mlsanity.com/paper/2604.22888/) (2026)
- [Poise: Position-Aware One-Instruction Skill Injection for Silent Execution on LLM Agents](https://mlsanity.com/paper/2606.07943/) (2026)

## Built on by (most cited first, 2 of 2)

- [SkillPoison: Progressive Skill Poisoning via Successful Experiences](https://mlsanity.com/paper/2610.07645/) (2026)
- [Evolutionary Safety of Recursive Self-Improving AI: Taxonomy, Risk Discovery, and Evaluation](https://mlsanity.com/paper/2609.31186/) (2026)

## Abstract (arXiv)

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 is the distinctive bottleneck: the target behavior must appear causally useful, recurrent, and generalizable before promotion. We evaluate four representative security-effect families using inert canary specifications. At 10% attacker support, across six mainstream LLM evolvers in SkillClaw, PoisonedEvolution embeds target behaviors in 546/600 trials (91.0% SER). On the structurally different Trace2Skill pipeline at the same ratio, it embeds target behaviors in 369/600 trials (61.5% SER), demonstrating transfer across evolution architectures. In a representative controlled study, three consistent attacker records suffice in a 30-record batch, whereas a single record is much weaker. Ablations identify recurring support, causal framing, and domain-aligned encoding as the main determinants of success. These findings expose evidence promotion as a security boundary for self-evolving agents.