Clear technical writing for AI agents.
Before: The retry subsystem shall initiate no more than three subsequent connection attempts within a sixty-second temporal window.
After: The retry subsystem must make no more than three additional connection attempts in 60 seconds.
STE helps an agent explain, rewrite, or review confusing technical content. It makes the meaning easier to follow without removing important details.
Install with the open skills CLI:
npx skills add phb123/ste --skill steAdd -g to make STE available across your projects. The installer supports Codex, Claude Code, Cursor, GitHub Copilot, and many other agents.
To install manually, copy skills/ste into your agent's skills directory. For example, Codex discovers global skills in ~/.codex/skills/.
STE activates when you express confusion or ask for clearer technical communication:
This explanation is confusing. What does eventual consistency mean?
Clarify this for a new engineer.
Simplify this procedure without changing its warnings or limits.
De-jargon this API description.
Review this runbook for clarity.
You can also invoke it directly:
Use $ste to rewrite this deployment note for an on-call engineer.
| Mode | Result |
|---|---|
| Explain | Answers first, defines necessary terms, and adds a concrete example when useful. |
| Rewrite | Returns clearer text while preserving facts, warnings, limits, and obligations. |
| Review | Identifies specific clarity problems and provides an improved version. |
The method favors direct sentences, one main idea per sentence, consistent terminology, explicit conditions, and short, scannable paragraphs.
Clearer must not mean less accurate. Before responding, STE checks that it preserved:
- Names, numbers, units, and limits.
- Warnings, conditions, exceptions, and sequence.
- Obligation strength such as must, shall, should, and may.
- Uncertainty and ambiguity in the source.
If the source does not identify an actor or meaning, STE exposes the ambiguity instead of inventing an answer.
See before-and-after examples for explanations, procedures, safety warnings, and reviews.
STE is a writing aid, not a grammar checker or certification tool. It does not activate only because a task is technical, and it does not replace subject-matter review for safety-critical content.
python -m unittest discover -s tests -v
uvx --from "git+https://github.com/agentskills/agentskills#subdirectory=skills-ref" skills-ref validate skills/steBehavioral cases are in evals/cases.yaml.
Issues and pull requests are welcome. Read CONTRIBUTING.md before changing behavior.