跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
搜索与研究

gza-explore-summarize

Summarize an explore task, extract the important findings, and suggest concrete next steps

mhawthorne · 已有 AI 审核记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add mhawthorne/gza --skill gza-explore-summarize

技能说明


name: gza-explore-summarize description: Summarize an explore task, extract the important findings, and suggest concrete next steps allowed-tools: Read, Bash(uv run gza show:), Bash(uv run gza log:), AskUserQuestion version: 1.0.0 public: true

Gza Explore Summarize

Summarize a completed explore task and turn its output into concrete next-step guidance. Use this when an explore task produced useful markdown or notes, but the user wants a faster handoff into planning, implementation, or a narrower follow-up exploration.

Inputs

  • Required: full prefixed explore task ID (for example, gza-1234)
  • Optional: context about what decision the user is trying to make next

If the user did not provide a full prefixed task ID, ask for it before proceeding.

Use the full prefixed task ID for all gza commands.

Goal

Produce a concise operator-facing synthesis:

  • what the explore task found
  • what matters most
  • what remains uncertain
  • what the best next steps are

This skill is not a raw inspection dump. Prefer synthesis and recommended actions over repeating the whole report.

Process

Step 1: Inspect the task

Run:

uv run gza show <TASK_ID>
uv run gza log <TASK_ID>

Use the output to verify:

  • task exists
  • task type is explore
  • current status
  • report or output content
  • any execution/log context that changes how confident the findings are

If the task is not found or is not an explore task, stop and explain the mismatch.

Step 2: Read the exploration output

Use the report content from gza show or the report file it references.

Extract:

  • main findings
  • evidence or examples that support those findings
  • recommendations already present in the report
  • unresolved unknowns, risks, or missing context

If the output is thin, ambiguous, or clearly incomplete, use AskUserQuestion to clarify what decision the user is trying to make from this research.

Step 3: Synthesize what matters

Summarize the exploration in a compact, decision-oriented way.

Focus on:

  • 3 to 7 findings at most
  • the implications of those findings
  • any disagreements, tradeoffs, or uncertainty that still affect the next step

Do not simply restate the report section by section.

Step 4: Recommend next actions

Suggest the most appropriate next step based on the exploration output:

  • create a plan task if the explore task narrowed the problem but design work is still needed
  • create an implement task if the path is already clear enough to execute
  • create another explore task if one key unknown still blocks good planning or implementation
  • create a review or docs-oriented follow-up only if the exploration clearly points there

When helpful, recommend concrete gza commands using the current task as lineage:

uv run gza add --type plan --based-on <TASK_ID> "..."
uv run gza add --type implement --based-on <TASK_ID> "..."
uv run gza add --type explore --based-on <TASK_ID> "..."

Prefer one recommended path plus one fallback, rather than a long menu of possibilities.

Step 5: Output format

Use this structure:

Explore Summary: Task <TASK_ID>

Task Context
- Status: <status>
- Prompt: <prompt>

Key Findings
- <finding>
- <finding>
- <finding>

What Matters
- <implication or decision-relevant takeaway>

Open Questions
- <remaining uncertainty>

Suggested Next Steps
1. <recommended next action>
2. <optional fallback or follow-up action>

Recommended Commands
- <command>
- <command>

Keep it concise. The goal is to help the user decide what to do next without rereading the whole exploration report.

Important notes

  • Use the explore output as primary evidence; do not invent conclusions the report did not support.
  • If the report contains clear recommendations, preserve them, but rewrite them more sharply when useful.
  • If the exploration does not support a confident implementation path, prefer recommending a plan task over jumping straight to implement.
  • If the user mainly wants raw task details, prefer /gza-task-info.
  • If the user wants a queue-level overview across many tasks, prefer /gza-summary.