AgentSkillsCN

prompt-schema-authoring

在此仓库中，按照所需的 `<role>`、`<task>`、`<input>` 和 `<output>` 结构，使用明确且与模型对齐的字段契约来创建或更新代理系统提示。适用于编辑 `app/prompts/*_system.md` 文件，或引入新的提示。

SKILL.md
--- frontmatter
name: prompt-schema-authoring
description: Create or update agent system prompts in this repository using the required `<role>`, `<task>`, `<input>`, and `<output>` structure with explicit, model-aligned field contracts. Use when editing `app/prompts/*_system.md` or introducing a new prompt.

prompt-schema-authoring

Use this skill to keep prompts explicit, contract-driven, and context-efficient.

Workflow

  1. •Identify the runtime contract before editing the prompt.
  • •Open the target prompt in */prompts/.
  • •Open the corresponding model in app/modules/*/models.py or any other schema we are using for the LLM input/output definitions
  1. •Write prompt sections in this order.
  • •<role>
  • •<task>
  • •<input>
  • •<output>
  • •Add <example_output> whenever possible.
  1. •Define <input> from real runtime payload only.
  • •List only fields actually passed in code.
  • •Add type and short purpose for each field.
  • •Exclude context-junk/internal fields unless the model truly needs them (for example path, chunk_index, chunk_count).
  1. •Define <output> from model contract only.
  • •State tool invocation requirements explicitly:
    • •pass a JSON object or JSON list depending on the tool definition
    • •return only the desired output
  • •Enumerate required and optional fields exactly as expected by the model.
  • •Explain field behavior clearly.
  • •Exclude internal/auto fields that should not come from the LLM (for example created_at).
  1. •

    Add one valid <example_output> that conforms to the model.

  2. •

    Keep contracts aligned end-to-end.

  • •If you change prompt output fields, update models, coercion/parsing, runtime logic, and tests in the same change.

Required Prompt Rules

  • •Keep instructions explicit and operational.
  • •Keep output contract field-by-field and typed.
  • •Avoid asking for wrappers/status/error fields unless the model requires them.
  • •Avoid asking for timestamps from the LLM.
  • •Avoid meta phrasing requirements that conflict with downstream synthesis.

Prompt Skeleton

md
<role>
...
</role>

<task>
...
</task>

<input>
Input is a JSON object with:
- `field_name` (type): purpose
</input>

<output>
You must call tool `tool_name` and pass a JSON object (not a JSON string).
Return only that tool call.

The tool argument must match `ModelName`:

- `field_name` (...)
  </output>

<example_output>
{
"field_name": "..."
}
</example_output>