dsh-reverse-skill

Maintaining a manual list of reverse engineering skills in DeepSeek Harness is tedious and error-prone. dsh-reverse-skill wraps all 87 SKILL.md files from the upstream reverse-skill project into a single Cordis plugin, distributed with the package and loaded automatically—no hand-written candidate lists required. The plugin is fully data-driven: it recursively scans skill directories, parses frontmatter, builds skill candidates, and registers a SkillProvider, so adding or removing a skill only requires changing the directory and the plugin syncs automatically.

The core value is a zero-maintenance integration of a community-validated security research skill set covering reverse engineering, penetration testing, and CTF scenarios. Models can invoke these skills through ctx.skills automatically, and users can also trigger them by name. The plugin includes repeatable verification to ensure all 87 skills are discoverable, uniquely named, and return non-empty content.

This plugin is ideal for developers and researchers working on authorized reverse engineering, penetration testing, and security research. If you already use DeepSeek Harness and want a scalable, trusted reverse engineering skill library without worrying about packaging, paths, or registration details, this is a great fit. Note that upstream agent definitions and some tool restrictions are not yet supported by the current dsh version, so you may need to handle those constraints at the harness level.