The Agent Skills Directory

[PROMPT_INJECTION]: The skill describes a methodology that involves ingesting and analyzing untrusted data from external sources, which presents an attack surface for indirect prompt injection.
Ingestion points: The agent is instructed to resolve DNS records (using dig) and fetch HTTP response content (using curl) from target subdomains (SKILL.md, Sections 2, 4, 5, and 8).
Boundary markers: The instructions lack explicit boundary markers or directions to treat the external data as non-executable text, potentially allowing malicious content in those responses to influence the agent's logic.
Capability inventory: The process involves executing various CLI tools including dig, curl, aws, heroku, and git based on the interpretation of the external data (SKILL.md, Sections 2, 4, and 5).
Sanitization: There are no provisions for sanitizing or validating the content received from external services before it is processed by the agent.
[COMMAND_EXECUTION]: The skill instructs the agent to execute various system commands and third-party CLI tools to perform detection and exploitation tasks.
Evidence: Procedures include commands such as dig, curl, aws s3 mb, heroku domains:add, and git (SKILL.md, Sections 2, 4, and 5).

subdomain-takeover