AgentSkillsCN

gap_resolver

阶段 1：分析部分场景数据，识别并补全缺失信息。适用场景：(1) 场景 JSON 中存在空值或待定字段；(2) 需要验证需求的完整性；(3) 需要向用户提出澄清性问题。在提示用户之前，系统会先尝试通过机器人选择器与 SKU 分析器自动解决相关问题。

SKILL.md
--- frontmatter
name: gap_resolver
description: "Stage 1: Analyzes partial scene data to identify and resolve missing information. Use when: (1) Scene JSON has null or TBD fields, (2) Need to validate completeness of requirements, (3) Need to generate clarifying questions for user. Attempts automatic resolution via robot-selector and sku-analyzer before prompting user."
metadata:
  stage: "1"

Gap Resolver

Overview

Smart gap analyzer that identifies missing data and categorizes by priority (critical/important/optional). Tells you what still needs resolution before proceeding.

When to Use This Skill

  • •After request_interpreter returns partial scene data with null/TBD fields
  • •User provides incomplete information
  • •Need to validate completeness before Stage 2

Workflow Instructions for Agent

Step 1: Save Scene Data to File Write the current partial_scene_data to a temporary JSON file:

python
import json
with open('temp_scene.json', 'w') as f:
    json.dump(partial_scene_data, f)

Step 2: Execute Script Run the gap analysis script:

bash
python skills/gap_resolver/scripts/resolve_gaps.py --scene_json "temp_scene.json"

Step 3: Parse JSON Output The script outputs:

json
{
  "gaps": [
    {
      "field": "robot_configuration.model_id",
      "type": "missing_robot",
      "priority": "critical",
      "description": "Robot model not specified",
      "has_default": false
    }
  ],
  "is_complete": false,
  "summary": {
    "critical_count": 1,
    "important_count": 2,
    "optional_count": 1
  }
}

Step 4: Formulate Natural Questions Based on the gaps, generate 8-10 conversational questions:

  • •Group by topic (robot, objects, workspace, task)
  • •Use natural language, not field names
  • •Example: "What robot model would you like to use?" (not "What is robot_configuration.model_id?")

Step 5: Ask User & Update Ask the questions conversationally, get answers, then update partial_scene_data with new information.

Step 6: Loop Until Complete Re-run this skill until is_complete: true

Step 7: Next Action Once complete → Load robot_selector skill

Usage

bash
python scripts/resolve_gaps.py --scene_json "path/to/partial_scene.json"

Gap Resolution Process

  • •Auto-Resolvable: Missing robot → robot-selector, Vague assets → sku-analyzer
  • •User Input Required: Missing positions, unclear task type, ambiguous constraints

Output

Returns list of gaps with resolution strategy and user questions for unresolved items

Input Schema

{ "partial_scene_data": "object" }

Output Schema

{ "is_complete": "boolean", "updated_scene_data": "object", "user_question": "string (Null if resolved internally, otherwise string)" }