Repository navigation
Releases: Yingye107/practical-work-kit
Release list
v0.2.1 — English-only presentation
Practical Work Kit now has an English-only public presentation across its README, installation guide, quick start, worked examples, plugin listing, and four skill interfaces.
- Use the same four shared skills in Codex and Claude: check work, challenge plans, shape ideas, and carry context.
- Read every example and starter prompt in English.
- Keep the original skill instructions and reference material; this release changes presentation rather than behavior.
- Earlier releases and Git history remain available.
Download practical-work-kit-0.2.1.zip and SHA256SUMS.txt. See the installation guide for either host.
Validation: 24 package tests passed. The 39-file ZIP passed source-byte comparisons and a scan for remaining Chinese characters. Claude Code 2.1.206 accepted both manifests in strict mode and installed the extracted ZIP in an isolated configuration, discovering all four skills and zero runtime components.
This verifies package contents and local skill discovery. Model responses and account-based Claude installation remain outside these checks. GitHub publishing is separate from public-directory approval by OpenAI or Anthropic.
v0.2.0 — Codex + Claude
Practical Work Kit now installs in both Codex and Claude from the same repository and release ZIP. Its four shared skills check work, challenge plans, shape ideas, and carry context into another conversation.
- Added Claude plugin and marketplace entry points, without copying or changing the shared skill instructions.
- Added English and Traditional Chinese instructions for Claude Code and Claude chat/Cowork.
- Retained the Codex entry point, licenses, source attribution, and prior releases.
For Claude Code, run:
claude plugin marketplace add Yingye107/practical-work-kit
claude plugin install practical-work-kit@practical-work-kitStart a new session and try /practical-work-kit:check-my-work. All four command names and Codex installation steps are in the installation guide.
In Claude chat or Cowork, use Customize → Plugins to add this GitHub marketplace or upload the ZIP, where those options are available to your account. See Claude's official guide.
Download practical-work-kit-0.2.0.zip for either host and SHA256SUMS.txt to verify it.
Validation: 24 package tests passed, including rejection of private files, secret signatures, unexpected runtime fields, version drift, and modified archives. The 40-file ZIP passed byte-for-byte checks. Claude Code 2.1.206 accepted both manifests in strict mode and installed all four skills from the extracted ZIP in an isolated configuration, with zero hooks, agents, MCP servers, or LSP servers.
This verifies packaging and local Claude Code discovery. It does not measure model answer quality or account-based Claude installation. GitHub availability is separate from public-directory approval by OpenAI or Anthropic.
Practical Work Kit 0.1.0 — 實用工作四助手
Practical Work Kit 0.1.0 / 實用工作四助手
Four reusable AI skills for checking work, challenging plans, shaping ideas, and handing off context.
Included
check-my-work— review an existing result for errors, omissions, and evidence.challenge-my-plan— test assumptions and suggest a practical alternative.shape-my-idea— turn an idea into a concrete direction, draft, or small experiment.carry-my-context— preserve decisions, drafts, progress, and the next step.- English / 繁體中文 guides, worked examples, source attribution, license files, and security policy.
Install
codex plugin marketplace add Yingye107/practical-work-kit --ref v0.1.0
codex plugin add practical-work-kit@practical-work-kitStart a new Codex conversation after installation. Use one source of these skills to avoid duplicate copies. See the repository README for compatible-host guidance.
The attached ZIP contains the installable skills-only package and its user documentation. It excludes source maintenance tools, tests, CI, private records, and Git history. SHA256SUMS.txt lets you verify the ZIP. The package adds no MCP server, runtime scripts, or background hooks.
AI instructions do not provide a sandbox or guarantee correctness. GitHub publication does not imply approval in the OpenAI public plugin directory. Maintainer checks are not a certification of all secret formats, legal compliance, or host behavior.
Licensed under Apache-2.0, with applicable upstream licenses and attribution retained. English README · 繁體中文 · Security policy.