What problem does it solve? New DeepSeek Harness (DSH) users often get stuck on startup failures, permission prompts, workspace confusion, or skills that silently fail to load. This Skill helps an assistant diagnose which layer is broken—model API, harness runtime, profile, or workspace—and provide verified, command-backed fixes instead of guesses. ## Core Features & Use Cases - Layered Diagnosis: Separates problems into four layers (model, harness, profile, workspace) so vague reports like "it doesn't respond" get narrowed down quickly. - Startup Troubleshooting: Covers the most common launch failures—port conflicts (EADDRINUSE on 3080), Node version issues, paths with spaces, slow first npx downloads, and missing API keys—with concrete commands like lsof -nP -iTCP:3080 and dsh --profile web --dump-config. - Skill Discovery Rules: Explains the SKILL.md frontmatter requirements, kebab-case key naming, and the ranked discovery paths (<project>/.dsh/skills through ~/.agents/skills) so users can fix skills that silently disappear. - Use Case: A user reports their new skill "isn't showing up." The Skill directs the assistant to check the frontmatter for camelCase keys like userInvocable instead of user-invocable, which causes the skill to be silently skipped. ## Quick Start Help me get DeepSeek Harness running for the first time and figure out why my skill is not being discovered.
