dsh-reverse-skill is a DeepSeek Harness plugin that packages reverse-engineering and security research skills for authorized penetration testing and security analysis.
The tool solves the problem of organizing and distributing a comprehensive collection of reverse-engineering techniques and knowledge by wrapping 88 skill definitions into a single Cordis plugin for the DeepSeek Harness framework. Rather than requiring manual maintenance of a candidate list, the skills are bundled with the plugin and automatically loaded when the plugin is activated. Users and AI models can invoke these skills either through automatic tool-calling or by name, depending on each skill's configuration.
Developers should install this tool directly from GitHub rather than npm, as the npm registry version is incompatible with current DeepSeek Harness releases and will be flagged as unsafe by the installer. The tool is designed for authorized security work only—reverse engineering, penetration testing, and security research on systems where the user has legal permission to operate. The project explicitly acknowledges that it contains dual-use content with legitimate security applications that could also be misused, and users are responsible for ensuring authorized use. Installation is straightforward via the dsh plugin command, after which the 88 skills are automatically registered and available to both the model and users.
The project maintains active compatibility with the current DeepSeek Harness release cycle, with the main branch always tracking compatible versions. Development activity shows responsiveness to framework changes, as evidenced by the explicit incompatibility warnings and version management between the GitHub and npm distributions. The codebase includes careful attention to resource path resolution in the plugin loader to ensure skills are correctly located after compilation.
