ste-zh skill
ste-zh is a skill for strict agent reports in Chinese: conclusion first, work status and verification status kept apart.
- 875 ★ on the repository
- Last commit
Minimal access
Why this level
- Changes only how replies are worded; files and commands are untouched.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/ste-zhDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- Changes only how replies are worded; files and commands are untouched.
This is third-party code. Review the repository files before installing.
What it does
ste-zh puts an agent's final reports into controlled Chinese modelled on ASD-STE100, the aviation standard for simplified technical English. The first line names the result and its scope, followed by the cause, evidence, a list of changes and checks. "Done" and "verified" are recorded separately, and anything unchecked is flagged openly. When a human must decide, the agent lays out numbered options with a recommendation, so the reply can be a single digit. The skill changes only how the agent talks: code, comments, commit messages and files stay as they were.
Who it is for. Chinese-speaking developers whose agent sends long reports full of guesses instead of verified facts. It also suits teams where reports are read without repository access.
Good fit when
- An agent finished a bug fix, but the report does not show what was run and what was not.
- A disputed point needs a quick decision: the agent offers options with a recommendation and a digit is enough.
- Reports after a long session are read on a phone and the conclusion must come first.
- A team agrees on one report format for several agents.
Not a fit when
- Reports and discussion are in Russian or English: the rules are written for Chinese.
- The strict format must also apply to code, comments or commit messages, which the skill leaves alone.
- A guarantee is needed that the agent never drifts from the format: nothing enforces it.
Example request
/ste Fix the token refresh race and report what changed, what was verified and what remains unverified.Limitations
The skill works only after explicit activation with `/ste` or a phrase such as "按 ASD-STE100 汇报". A new session needs it switched on again. The format holds only because the agent follows context instructions; there is no programmatic check. In very long sessions and after context compaction the agent may drift, and `/ste` is then sent again. The rules are built for Chinese, with terminology in a separate `references/terminology.md`. Traditional characters have their own repository, ste-zht.
How to disable. Send the agent "stop ste" or "停止 ste". To remove it for good, run `npx skills remove --global ste` or delete the `ste` folder from the skills directory, along with the activation line in `CLAUDE.md` if you added one.
FAQ
How do I turn the skill on?
Send `/ste` or the phrase "用 STE 规范输出". Until then it has no effect, even when installed.
What changes in a report?
The first line names the result. Then come the cause, evidence with file and line references, the changes, and checks split into verified and unverified.
Does the skill touch code?
No. It governs only how the agent speaks in reports and discussion. Code, comments, commits and created files stay unchanged.
How do I enable it in every session?
Add a line about activating ste at the start of every session to your global rules, for example `~/.claude/CLAUDE.md`. The README counts such a line as explicit activation.
What is the license?
MIT. A separate repository, ste-zht, covers traditional characters.
Discussion
No questions or reviews yet.
Related
SkillsOpen-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
An engineering system for coding agents: 71 agents, 311 skills, hooks, memory, rules and the AgentShield security scanner
DeepSeek Harness is an open-source AI agent from DeepSeek where every feature is a plugin. It runs in the browser, on macOS and on Windows.
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Ollama runs open models on your own machine and serves them to agents through a local API compatible with OpenAI and Anthropic.
README in short
The README compares a typical wordy bug report with one written under ste-zh rules. It covers installation through the `skills` CLI and by hand, activation and leaving the mode.