dsh-claude-skills

Juggling engineering specs, product docs, marketing copy, compliance reviews, and deep research inside DeepSeek Harness usually means rewriting prompts, bolting on scripts, and managing context for every domain. dsh-claude-skills ports 362 battle-tested Claude Code skills verbatim into the Harness global skill layer, turning 18 specialised workflows from code generation to grant writing into single-utterance, reusable capabilities and removing the overhead of rebuilding prompts and glue code each time.\n\nEvery skill ships as a SKILL.md bundle (narrative, scripts, references, assets) discovered through the standard Harness registration pipeline, backed by 658 Python utilities (predominantly stdlib-only) that span Markdown-to-HTML rendering, literature review, patent search, AEO optimisation, full C-suite advisory, and deep-work routines such as capture, reflect, handoff, deep-work, and meetings. A built-in Skill Security Auditor scans each skill for command injection, data exfiltration, and prompt-injection vectors before installation, adding a safety gate when batch-introducing third-party content.\n\nIt is ideal for developers, product managers, marketing and QA teams, researchers, and independent consultants who already work inside DeepSeek Harness (Web, TUI, or headless) and want to chain multi-domain expertise in a conversation. After installation, all 359 fully usable skills land in the global skill layer, and a natural-language task description automatically matches the right one no memorising skill names required.