Aegis
GanyuanRan
Software-engineering method pack for coding agents, with skills for baseline-first planning, systematic debugging, prompt hygiene, verification before completion, and repair/retirement tracking.
biedongbin/dsh-claude-compat
DSH plugin: bridge Claude Code's .claude/ directory (skills, commands, rules) into DeepSeek Harness natively
Install
via npm
Resolves a published tarball rather than cloning the repository, and installs without any extra setup. Swap `web` for your profile name if you run another one.
via GitHub · npm package
Installing from GitHub runs the project's build script, which pnpm blocks until you allowlist it — run the command once and pnpm prints the exact key to add under `allowBuilds` in ~/.dsh/profiles/web/pnpm-workspace.yaml.
Due diligence

dsh-claude-compat is a DeepSeek Harness plugin that makes Claude Code conventions available from a project’s .claude/ directory and the user-level ~/.claude/ directory. It registers skills/**/SKILL.md and commands/*.md through a DSH skill provider: names and descriptions appear in the model-visible catalogue, while bodies load on demand through the skill tool. Markdown command files are also available as slash commands such as /command-name. Files in rules/ use a different path: their contents are concatenated, wrapped in <system-reminder>, and prepended as a user-role message to the message array once per session.
The plugin reads these files; the README does not describe writing or modifying them. Project entries take precedence over DSH-native and ~/.claude/ entries, while DSH-native entries take precedence over user-level copies. Project rules are cached per session working directory, and changes take effect in the next session. enableSkills, enableRules, rulesMaxBytes, userRulesMaxBytes and project-root discovery can be configured.
dsh-claude-compat is intended for projects that already use .claude/skills/, .claude/commands/ or .claude/rules/, and requires DSH with a profile. It is a poor fit when these conventions are not used, or when DSH’s built-in dsh-agent-instructions is sufficient for CLAUDE.md and AGENTS.md, which this plugin deliberately does not touch. Its terminal-surface risk is its access to project and home-directory .claude files: rules are injected verbatim as model instructions, so only trusted content should be placed there.
dsh-claude-compat discovers a project root using the markers in projectRootMarkers, whose default is [".git"]. It reads .claude/skills/**/SKILL.md, .claude/commands/*.md, and .claude/rules/*.md, plus the corresponding paths under userClaudeDir, defaulting to ~/.claude.
Skills and commands are exposed through the DSH skill provider. A skill’s name and description enter the model-visible catalogue, and its body is fetched when the model uses the skill tool. Command files also appear in the slash menu. Nested skill directories are flattened, such as gitnexus/gitnexus-guide becoming gitnexus-gitnexus-guide; invalid frontmatter names fall back to the directory name. DSH requires kebab-case skill names.
Rules are concatenated and injected verbatim inside a <system-reminder> envelope as a user-role message at the front of the message array once per session. Project and user rules with the same basename are deduplicated in favour of the project file. CLAUDE.md and AGENTS.md are not read by this plugin.
| Key | Default | Purpose |
|---|---|---|
enableSkills |
true |
Enables the skills and commands provider |
enableRules |
true |
Enables rule injection |
rulesMaxBytes |
65536 |
Caps project rule text |
userRulesMaxBytes |
65536 |
Caps ~/.claude/rules text |
projectRootMarkers |
[".git"] |
Project-root discovery markers |
skillRank |
50 |
Rank for project skills |
skillSource |
project-claude |
Project catalogue source tag |
userSkillRank |
700 |
Rank for user skills |
userSkillSource |
user-claude |
User catalogue source tag |
userClaudeDir |
~/.claude |
User-level Claude directory |
Project entries use rank 50, DSH bundled skills use 600, and user entries use 700; lower rank wins.
The README requires DSH with a profile, such as web, and a project using Claude Code directory conventions. All three directory types are optional. Skills are reconciled into the catalogue without a restart, while rules are read for each new session and cached per session working directory. Editing a rule during a session therefore affects the next session.
Written from the project's own documentation and kept in sync with it. Where the two disagree, the source is authoritative — read the README on GitHub
Same category
GanyuanRan
Software-engineering method pack for coding agents, with skills for baseline-first planning, systematic debugging, prompt hygiene, verification before completion, and repair/retirement tracking.
superdesigndev
Design skill for UI and marketing graphics on the Superdesign canvas: reads the repo for context, extracts its design system, then generates and iterates branchable design drafts, flow pages, and reusable components through the Superdesign CLI.
Ikalus1988
Failure-recovery memory: search and record failure-recovery lessons from real engineering sessions, with BM25 + semantic RAG retrieval and a lessons knowledge base.
Vladimir-Human
Cleans AI traces from Russian text: finds chatbot copy-paste artefacts (ChatGPT, Gemini, Grok, Perplexity, DeepSeek) and rewrites the text into natural prose on request; 39 regex markers with an evidence registry, offline, text-only bundle.
No one-line install — this plugin lives inside a larger repository and publishes no npm package.
dhicoc
Complete reverse-skill pack (85 SKILL.md) as a DeepSeek Harness Cordis plugin: reverse engineering, authorized pentesting and security-research skill router.
zhaiyateng
Design-aesthetics skill pack (10 styles: dark SaaS, minimal white, neumorphism, brutalism, glassmorphism, Japanese minimal, bento grid, cyberpunk, vaporwave, art deco) with runnable landing-page demos: tokens, component rules, forbidden lists, and acceptance checklists per style.