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.
bash
python3 /Users/fox/.openclaw/workspace/skills/neovim-daily-digest/scripts/neovim_digest.py --json --limit 20
top.rss?t=day for the true daily leaderboardnew.rss and hot.rss to catch later US-time posts and emerging discussionsbash
python3 /Users/fox/.openclaw/workspace/skills/neovim-daily-digest/scripts/neovim_digest.py --json --limit 20 --week-backfill
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.Keep these by default:
:keepalt, buffers, LSP, Tree-sitter, cmp, diff, config/workflow)Drop or strongly downrank these by default:
Read references/filtering.md only when you need to explain or tune the heuristics.
Default Markdown output:
bash
python3 /Users/fox/.openclaw/workspace/skills/neovim-daily-digest/scripts/neovim_digest.py --limit 10
JSON for agent-side rewriting:
bash
python3 /Users/fox/.openclaw/workspace/skills/neovim-daily-digest/scripts/neovim_digest.py --json --limit 20
Include lower-priority near-misses:
bash
python3 /Users/fox/.openclaw/workspace/skills/neovim-daily-digest/scripts/neovim_digest.py --limit 20 --include-near-misses
Prefer a digest that is useful to a working Neovim user, not a full feed dump.
Good output:
Bad output: