Neovim Daily Digest
Use scripts/neovim_digest.py as the primary data source. It fetches Reddit RSS directly, which is more reliable than browsing normal Reddit pages in environments that trigger Reddit's anti-bot wall.
Workflow
- 1. Run the script in JSON mode first:
CODEBLOCK0
- 2. Keep the highest-scoring items by default. The script already combines:
-
top.rss?t=day for the true daily leaderboard
-
new.rss and
hot.rss to catch later US-time posts and emerging discussions
- 3. If the user explicitly asks for more coverage than the current day supports, rerun with week backfill:
CODEBLOCK1
- 4. Rewrite the output into concise Markdown in the user's language:
- 1 short intro line explaining the scope
- 1 short observations section if useful
- bullet list of selected posts with title link and 1-sentence takeaway
- optional GitHub link when the post clearly links to a repo/plugin
- 5. Be explicit when daily
top/day is thin. If only a handful of posts exist there, say that new and hot were used as supplemental feeds rather than pretending you found a full 20 in top/day.
Filtering Rules
Keep these by default:
- - concrete Neovim usage tips (
:keepalt, buffers, LSP, Tree-sitter, cmp, diff, config/workflow) - plugin releases, plugin updates, and new plugin launches
- AI coding workflow posts when they ship concrete Neovim functionality
- troubleshooting threads only when they expose a broadly useful debugging/config lesson
Drop or strongly downrank these by default:
- - weekly/monthly sticky threads
- pure screenshots or setup flex posts with no actionable detail
- generic theme/colorscheme aesthetics unless there is a meaningful feature update
- vague recommendation bait or support noise with little reusable value
Read references/filtering.md only when you need to explain or tune the heuristics.
Useful Script Modes
Default Markdown output:
CODEBLOCK2
JSON for agent-side rewriting:
CODEBLOCK3
Include lower-priority near-misses:
CODEBLOCK4
Output Expectations
Prefer a digest that is useful to a working Neovim user, not a full feed dump.
Good output:
- - highlights the few posts worth opening
- says why each post matters
- links directly to the Reddit post
- includes repo links when obvious
Bad output:
- - repeats every post mechanically
- treats sticky threads as news
- overweights pure aesthetics
- claims there were 20 strong posts when there were only 7-10
技能名称：neovim-daily-digest
详细描述：
Neovim 每日摘要
使用 scripts/neovim_digest.py 作为主要数据源。它直接获取 Reddit RSS，这比在触发 Reddit 反爬虫墙的环境中浏览普通 Reddit 页面更可靠。
工作流程
- 1. 首先以 JSON 模式运行脚本：
bash
python3 /Users/fox/.openclaw/workspace/skills/neovim-daily-digest/scripts/neovim_digest.py --json --limit 20
- 2. 默认保留得分最高的条目。该脚本已整合：
- top.rss?t=day 获取真实的每日排行榜
- new.rss 和 hot.rss 捕捉后续美国时区的帖子和新兴讨论
- 3. 如果用户明确要求覆盖比当天更多的内容，则使用周回填重新运行：
bash
python3 /Users/fox/.openclaw/workspace/skills/neovim-daily-digest/scripts/neovim_digest.py --json --limit 20 --week-backfill
- 4. 将输出重写为用户语言的简洁 Markdown 格式：
- 1 行简短介绍，说明范围
- 1 行简短观察部分（如有用）
- 选定帖子的项目符号列表，包含标题链接和一句话要点
- 当帖子明确链接到仓库/插件时，添加可选的 GitHub 链接
- 5. 当每日 top/day 内容稀少时，要明确说明。如果只有少量帖子存在，则说明 new 和 hot 被用作补充源，而不是假装在 top/day 中找到了完整的 20 条。
过滤规则
默认保留以下内容：
- - 具体的 Neovim 使用技巧（:keepalt、缓冲区、LSP、Tree-sitter、cmp、diff、配置/工作流程）
- 插件发布、插件更新和新插件发布
- 提供具体 Neovim 功能的 AI 编码工作流程帖子
- 仅当暴露了广泛有用的调试/配置经验时的故障排除线程
默认丢弃或强烈降级以下内容：
- - 每周/每月的置顶帖
- 纯截图或设置展示帖，无可行细节
- 通用主题/配色方案美学，除非有有意义的特性更新
- 模糊的推荐诱饵或支持噪音，无重复使用价值
仅当需要解释或调整启发式规则时，才阅读 references/filtering.md。
有用的脚本模式
默认 Markdown 输出：
bash
python3 /Users/fox/.openclaw/workspace/skills/neovim-daily-digest/scripts/neovim_digest.py --limit 10
用于代理端重写的 JSON：
bash
python3 /Users/fox/.openclaw/workspace/skills/neovim-daily-digest/scripts/neovim_digest.py --json --limit 20
包含低优先级的近似匹配：
bash
python3 /Users/fox/.openclaw/workspace/skills/neovim-daily-digest/scripts/neovim_digest.py --limit 20 --include-near-misses
输出期望
优先提供对工作中的 Neovim 用户有用的摘要，而不是完整的源转储。
好的输出：
- - 突出显示值得打开的少数帖子
- 说明每个帖子的重要性
- 直接链接到 Reddit 帖子
- 在明显时包含仓库链接
差的输出：
- - 机械地重复每个帖子
- 将置顶帖视为新闻
- 过度强调纯美学
- 声称有 20 个强帖子，但实际上只有 7-10 个
