The Agent Skills Directory

[SAFE]: The content of both SKILL.md and CACHE_POISONING_TECHNIQUES.md is strictly educational and instructional. It describes common web security vulnerabilities and provides manual testing steps for security researchers.
[COMMAND_EXECUTION]: The skill includes several example curl commands for testing HTTP headers and path confusion. These are presented as snippets for a user or agent to use in a controlled security auditing environment, not as automated or obfuscated malicious actions.
[INDIRECT_PROMPT_INJECTION]: The playbook instructs the agent on how to interact with and analyze external web resources. While this ingestion of external data constitutes a potential attack surface for indirect prompt injection, the skill is a reference guide intended for security analysis and includes sections on defense and sanitization.

web-cache-deception