Install focused Agent Skills from one trusted source, with complete payloads and independently verified releases.
简体中文 · Browse the catalog · Contribute
Zhijian Skills is the canonical source for nineteen focused Agent Skills spanning Codex operations and experience, workflow orchestration, model reasoning and infrastructure, knowledge systems, research, information design, and publishing.
List all nineteen Skills:
npx skills add zjp1997720/zhijian-skills --listInstall only what you need:
npx skills add zjp1997720/zhijian-skills --skill wechat-stylerInstall globally for a specific Harness:
npx skills add zjp1997720/zhijian-skills \
--skill codex-model-routing-team --agent codex --global --copy --yesThis is the only publishing repository. New Skills, releases, Issues, and contributions all belong here.
Open Experts · Skills · Connectors → Skills → Bundles → Add marketplace and enter:
zjp1997720/zhijian-skills
The full Git URL https://github.com/zjp1997720/zhijian-skills.git also works. Select zhijian-skills after adding it and install individual bundles as needed. Adding the marketplace does not install every Skill.
Use the repository URL, not a /tree/main/skills web page or raw JSON URL. Git distribution preserves each complete Skill payload, including scripts, references, and assets.
Catalog availability does not imply WorkBuddy runtime compatibility. Entries marked 需 Codex 宿主能力 require Codex host capabilities. Other entries may still require local CLIs, authentication, or dependencies; consult their documentation. This adapter does not change those requirements.
Maintainers: the catalog is generated from registry/skills.json and existing Chinese documentation. Run python3 scripts/build_workbuddy_marketplace.py after catalog/version changes; CI checks for drift.
The manifests and repository-local symlinks under plugins/ are generated; editable Skill source remains under skills/. Installation materializes the links into complete files. Git checkouts must support symlinks. Verified with macOS WorkBuddy; Windows has not been tested. Run python3 scripts/verify_workbuddy_install.py --cli <WorkBuddy-bundled-codebuddy> to check installation, cold-start discovery, and payload hashes in a temporary config. Add --source zjp1997720/zhijian-skills to verify the public source.
| Area | Skill | Result | Documentation |
|---|---|---|---|
| Model infrastructure | codex-cli-model-bridge |
Add verified loopback models to Codex without hiding ChatGPT history | Docs |
| Codex control | codex-doctor |
Diagnose context, configuration, and workspace drift without changing files | Docs |
| Cross-agent handoff | codex-external-handoff |
Launch and supervise persistent Codex App Server threads from WorkBuddy or Claude Code | Docs |
| Codex control | codex-handoff |
Continue an oversized or slow Codex task in a fresh task with compact context | Docs |
| Image generation | codex-image-gen |
Reuse the logged-in Codex CLI OAuth state to generate and edit images without an API key | Docs |
| Codex control | codex-model-routing-team |
Compile parallel work into TeamPlans; use Sol Medium execution and risk-based routing | Docs |
| Codex control | codex-skill-admin |
Audit, disable, restore, and verify local Codex Skills | Docs |
| Codex experience | codex-theme-studio |
Build reversible macOS Codex themes from safe variables, custom artwork, or bundled presets | Docs |
| Knowledge systems | enterprise-clone-builder |
Build a structured enterprise digital-twin repository from evidence | Docs |
| Model reasoning | gpt56-sol-pro-consult |
Get a file-grounded, model-verified GPT 5.6 Sol Pro second opinion through Codex Chrome | Docs |
| Information design | html-express |
Turn dense material into a clear, self-contained HTML report | Docs |
| Long-form writing | leadbook |
Produce evidence-backed Chinese business books and white papers with auditable quality gates | Docs |
| Workflow orchestration | light-plan-and-work |
Plan bounded work briefly, execute immediately, and escalate only on heavy conditions | Docs |
| Release governance | skill-open-sourcer |
Audit, package, document, verify, and publish Agent Skills | Docs |
| Content research | wechat-article-search |
Discover WeChat public-account articles as structured JSON | Docs |
| Editorial publishing | wechat-styler |
Convert Markdown into polished, WeChat-compatible inline HTML | Docs |
| Content archiving | web-clipper |
Save public article URLs and bounded archive pages as structured Markdown | Docs |
| Model infrastructure | workbuddy-cli-model-bridge |
Connect verified CLI subscription models to WorkBuddy through a loopback proxy | Docs |
| Content archiving | wxmp-article-harvester |
Export a public WeChat account into verified Markdown, indexes, and a completion report | Docs |
- One editable source. Every public Skill is maintained on
mainin this repository. - Complete installation units. Supporting scripts, references, themes, and assets travel with each Skill.
- Independent versions, one repository. Every Skill owns its version, Changelog, canonical Tag, and tests while sharing this publishing source.
codex-model-routing-team can be invoked explicitly. Its documentation also includes an optional AGENTS.md authorization block for automatic activation when parallel execution has a clear net benefit.
skills/<name>/ complete agent-facing install payload
docs/skills/<name>/ human-facing English and Chinese documentation
registry/skills.json versions, validation, capabilities, and Harness support
assets/readme/ Portfolio identity assets
Every install and release resolves through this repository. The Portfolio does not create or synchronize standalone Skill repositories.
Read CONTRIBUTING.md before opening an Issue or pull request. The Portfolio is released under the MIT License; bundled Skill notices remain with their respective payloads.
