The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous ripgrep (rg), ast-grep, and semgrep patterns for scanning project source code to identify potentially vulnerable sinks and insecure configurations (SKILL.md).
[PROMPT_INJECTION]: The skill describes an attack surface for indirect prompt injection as it instructs the agent to ingest and analyze untrusted project data.
Ingestion points: Project source code and configuration files analyzed via auditing tools (SKILL.md).
Boundary markers: None explicitly defined in the instruction set.
Capability inventory: Execution of filesystem search and static analysis commands via rg, ast-grep, and semgrep.
Sanitization: No specific sanitization or filtering of tool output is prescribed in the hunting methodology.
[EXTERNAL_DOWNLOADS]: The skill references authoritative security research websites, official vulnerability databases, and verified security advisories (e.g., GitHub Security Advisories, Snyk, PortSwigger, HackerOne) to provide educational context and exploitation examples.
[SAFE]: The skill includes various XSS payloads, including Base64-encoded canaries (e.g., atob('YWxlcnQoMSk=') which decodes to alert(1)), for the purpose of security testing and validation. These are intended for use against target applications and do not pose a threat to the agent host.

hunt-xss