The Agent Skills Directory

[DYNAMIC_EXECUTION]: The scripts/openghost.sh script employs a Base64 encoding pattern combined with Python's exec() function to pass and execute inline scripts within the Docker container. This is used to facilitate stable execution of Python code in the sandboxed environment.
[INDIRECT_PROMPT_INJECTION]: As a penetration testing tool, the skill ingests untrusted data from target web applications. This content is processed and summarized for the agent, creating a potential surface where malicious target responses could influence the agent's reasoning.
[EXTERNAL_DOWNLOADS]: The skill's operation involves pulling the ghcr.io/vaibhavsing/openghost-sandbox:latest Docker image, which is an external resource provided by the author to support the security toolchain.
[COMMAND_EXECUTION]: The orchestration layer utilizes subprocess.run and shell commands to manage the Docker lifecycle and execute allowlisted security tools like nmap, sqlmap, and nuclei.

openghost-skill