Users are injecting adversarial prompts into our chatbot that corrupt the agent's behavior for the rest of the session. Even after the malicious message, the agent keeps behaving incorrectly. What's happening and how do I prevent it? - [ ] Identifies context poisoning as degradation pattern - [ ] Explains how malicious content persists in context window - [ ] Recommends input validation/sanitization - [ ] Suggests context isolation or session segmentation - [ ] May recommend masking or filter...
Scanned 9/20/2026
Install to Claude Code
Skills Directorynpx -y skills add tstapler/dotfiles --skill edge-case-context-poisoning --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Edge Case Context Poisoning?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/tstapler-edge-case-context-poisoning)More formats (shields.io, HTML) on the badges page.
---
name: edge-case-context-poisoning
description: ''
---
# Prompt
Users are injecting adversarial prompts into our chatbot that corrupt the agent's behavior for the rest of the session. Even after the malicious message, the agent keeps behaving incorrectly. What's happening and how do I prevent it?
# Expected
- [ ] Identifies context poisoning as degradation pattern
- [ ] Explains how malicious content persists in context window
- [ ] Recommends input validation/sanitization
- [ ] Suggests context isolation or session segmentation
- [ ] May recommend masking or filtering strategies
- [ ] References context-degradation.md for poisoning patternsIs this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!
