classify-structured
GitHub基于JSON Schema的多维度文本分类技能。支持按指定轴（如情感、主题）及标签集同时分类，使用zerogpu工具执行，通过heredoc确保原文本解析安全。
Trigger Scenarios
Install
npx skills add zerogpu/zerogpu-router --skill classify-structured -g -y
SKILL.md
Frontmatter
{
"name": "classify-structured",
"description": "Multi-axis classification using a JSON schema mapping categories to allowed labels (gliner2-base-v1). Use when the user wants to classify text along several dimensions at once, e.g. \"classify by sentiment and topic\" with explicit label sets per axis.",
"allowed-tools": "Bash(zerogpu classify_structured*)",
"argument-hint": "<text> -s '<json schema>'"
}
Run schema-driven classification:
zerogpu classify_structured $ARGUMENTS
Quoting (required, to survive shell parsing of arbitrary user text): format $ARGUMENTS with the source text wrapped via heredoc command substitution, then flags after. Inside the heredoc, paste the user's text verbatim — do not escape:
"$(cat <<'ZGPU_T'
<the source text, verbatim, multi-line and special chars all OK>
ZGPU_T
)" -s '{"sentiment":["positive","negative","neutral"],"topic":["support","billing","product"]}'
Schema is a single-quoted JSON object mapping each axis to its allowed labels. Output is a JSON object with one chosen label per category.
Savings note: only if the command output literally contains a line starting with 💰 ZeroGPU savings, append that exact line, unchanged, as the last line of your reply. If no such line is present, say nothing about savings and do not mention or suggest /zerogpu-router:cost-savings — this note is intentionally occasional, not shown every time.
Version History
- 7b9b503 Current 2026-07-24 11:47
