By name: summarize-children trigger: /summarize-children description: Analyze and synthesize results from multiple child instances parameters:
- name: children required: true
- name: format default: "structured"
Summarize Children Skill
When triggered, analyze results from child instances and synthesize into coherent summary.
Usage
/summarize-children [format]
Behavior
- Retrieve structured results from all child instances; prefer child summaries/artifacts over raw transcript output
- Identify common themes and conflicts
- Extract key findings and decisions
- Synthesize into requested format
- Highlight critical items requiring attention
If a separate synthesis child is needed, spawn it with a non-Claude provider (gemini, codex, copilot, or cursor) unless the user explicitly requested Claude. Do not paste large raw child transcripts into your own context.
Parameters
- children: List of child instance IDs or "all" for all children
- format: Output format (structured, executive, technical, markdown)
Example
/summarize-children structured
This would:
- Collect all child results
- Categorize findings by type (errors, recommendations, decisions)
- Identify consensus and disagreements
- Produce structured summary
Formats
Structured
- Grouped by finding type
- Severity-sorted
- File references included
- Action items highlighted
Executive
- High-level overview
- Critical items only
- Business impact focus
- Recommendations prioritized
Technical
- Detailed technical findings
- Code snippets included
- Architecture implications
- Implementation guidance
Markdown
- Portable markdown document
- Suitable for documentation
- Section-organized
- Links to source files
Output
A comprehensive summary containing:
- Overview of child tasks completed
- Aggregated findings by severity
- Consensus recommendations
- Conflicting viewpoints (if any)
- Next steps and action items
- Metrics (time spent, LOC analyzed, issues found)