export-reply（导出回复）

Export Reply 💾

Save any agent reply or full conversation to a local file. Remembers your last export settings — repeat exports take one confirmation.

Highlights

Usage

Just say 保存 / 导出 / save / export. The agent asks all options in one message and infers what you already stated. Next time: "Use same settings as last time? [yes / change]"

Install

clawhub install export-reply

Optional — better PDF / DOCX support:

pip3 install fpdf2 python-docx markdown

<!-- Agent instructions below. Not rendered on ClawHub. -->

Interaction Flow (MUST follow exactly)

Step 0 — Check saved preferences

Before asking anything, run:

python3 skills/export-reply/scripts/prefs.py --action get

Branch A — Preferences exist (output is JSON, not null): Show a single message:

上次你用的是：{scope} · {mode} · {format} → {path}（{saved_at}）
直接沿用这个方式，还是重新选？[好/是 = 沿用 | 改/新 = 重选]

• User confirms (好/是/沿用/yes/ok or empty reply) → skip to Step 2 with saved prefs
• User wants to change → show Step 1 prompt

Branch B — No preferences (output is null): Go directly to Step 1.

Step 1 — Ask all options in one message

Combine only the unknowns into one single message:

请确认导出选项：
📄 内容范围：仅当前回答 / 完整对话？
✂️ 内容模式：原文保留 / 精简摘要？
📁 格式：MD · TXT · HTML · PDF · DOCX · 全部？
💾 保存路径：默认 ~/Desktop/ 还是自定义？

Infer from trigger phrase — skip already-known fields:

Step 2 — Execute export

1. Write content to temp file, then run:

python3 skills/export-reply/scripts/summarize_conversation.py \
  --input /tmp/export_raw.md \
  --output /tmp/export_staged.md \
  --mode {raw|summary} \
  --title "{TITLE}"

python3 skills/export-reply/scripts/export_reply.py \
  --file /tmp/export_staged.md \
  --format {fmt} \
  --output {path} \
  --title "{TITLE}"

1. On success, save preferences:

python3 skills/export-reply/scripts/prefs.py --action set \
  --scope {full|reply} --mode {raw|summary} \
  --format {fmt} --path {path}

1. Reply:

✅ 已保存：{full_path}（{size} · {format}）

Step 3 — Error handling

Content Modes

Raw (原文保留)

Verbatim, role-labeled:

## 用户
{user message}

## 助手
{agent reply}

Pass --mode raw.

Summary (精简摘要)

Bilingual condensed — every header and prose block must be Chinese + English.

• Remove: greetings, filler, repeated clarifications, narration
• Keep: core question, conclusions, code, decisions, action items
• Headers (exact):

- ## 核心问题 / Core Question - ## 结论 / Conclusions - ## 代码 / Code - ## 行动项 / Action Items - ## 关键决策 / Key Decisions

• Prose: Chinese first, English below
• Code: keep as-is, add bilingual caption above

Pass --mode summary.

Format Quick Reference

Common Triggers → Inferred Defaults

Preferences Script Reference

python3 skills/export-reply/scripts/prefs.py --action get    # read
python3 skills/export-reply/scripts/prefs.py --action set \
  --scope full --mode summary --format pdf --path ~/Desktop/  # save
python3 skills/export-reply/scripts/prefs.py --action clear  # reset

Stored at ~/.export_reply_prefs.json.

Related Skills

• self-improving — Persist agent learnings permanently
• ontology — Store structured knowledge from conversations