The Agent Skills Directory

Dynamic Script Creation and Execution: The skill is designed to write and run reproduction scripts (e.g., Python or C) to confirm the existence of security vulnerabilities. While executing dynamically generated code involves inherent security considerations, the skill mandates that all such activities occur within strictly isolated sandboxes or containers, which is a standard safety practice for vulnerability research.
Mandatory Host Isolation: The instructions explicitly prohibit the execution of commands directly on the host terminal. By requiring the use of emulation tools like QEMU or Unicorn, the skill ensures that potentially unstable or risky code is confined to a controlled environment, preventing side effects on the user's primary system.
Automated Workspace Data Processing: The skill ingests findings from JSON files within the local workspace to drive the reproduction logic. This integration allows for automated bug verification while operating within the established boundaries of the security testing environment.
Subagent Coordination: The skill describes the use of concurrent subagents to explore different logical approaches for bug reproduction. This orchestrated behavior is intended to improve the efficiency of the verification process and does not introduce additional security risks beyond the established execution constraints.

mantis-reproduce