| name | rss-reader |
| description | Fetch and manage RSS/Atom feed subscriptions stored in OPML format. Use when the user wants to list RSS subscriptions, add/remove RSS feeds, fetch and read latest articles, search across feeds, get a daily digest, or export feed content in markdown or JSON. Triggers on "rss", "rss feeds", "subscriptions", "fetch articles", "news feeds", "atom feed", "opml", "what's new", "digest", "search feeds", or any request to read, browse, search, or manage RSS content. |
RSS Reader
Manage RSS/Atom subscriptions (OPML) and fetch feed content as markdown or JSON.
Invocation
The rss CLI tool must be installed on PATH. If rss is not found (command not found error), stop and tell the user the tool is not installed.
Do NOT attempt any fallback. The tool must be installed before this skill can be used.
Default output is readable markdown. Add --json before the subcommand for structured JSON.
Configuration
- OPML file:
~/.config/llm/rss-reader.opml (auto-created on first use)
- Cache dir:
~/.cache/rss-reader/ (cached XML, 30-min TTL)
- Override via env:
RSS_OPML_PATH, RSS_CACHE_DIR
Commands
list
List all RSS subscriptions, grouped by category.
rss list
add / remove
rss add <url> [--title TITLE] [--category CATEGORY]
rss remove <query>
add is a no-op if the URL already exists. remove matches by title or URL substring.
merge
Import feeds from another OPML file. Deduplicates by xmlUrl.
rss merge <file>
recategorize
Move a feed to a different category.
rss recategorize <query> --category CATEGORY
fetch
Fetch and display feed content with full article content by default.
rss fetch <query> [--limit N] [--refresh] [--index N]
rss fetch --all [--limit N] [--refresh]
| Flag | Description |
|---|
--all | Fetch all subscribed feeds |
--limit N | Max items per feed (default: 10) |
--refresh | Bypass cache and re-fetch from source |
--index N | Show a specific item by 1-based index |
search
Search articles across ALL subscribed feeds by keyword.
rss search <keyword> [--limit N] [--refresh] [--feeds "rust,simon"]
Matches against title, description, and full content. Results sorted by date (newest first). Default limit: 20.
Cache-first: scans already-cached XML instantly. Only fetches uncached feeds (with --refresh, re-fetches all).
digest
Show recent articles across all feeds within a time window.
rss digest [--since 24h] [--limit N] [--refresh] [--feeds "rust,simon"]
| Flag | Description |
|---|
--since | Time window: 24h, 7d, 30m, etc. (default: 24h) |
--limit N | Max results (default: 50) |
--feeds | Comma-separated filter by feed title/URL substring |
Results grouped by feed, sorted newest first. Cache-first like search.
Agent Workflows
Answering "what's new?"
rss digest --since 24h
Then summarize the output for the user, highlighting items relevant to their interests.
Finding information on a topic
rss search "rust" --limit 10
Scan results, then use rss fetch <feed> --index N to drill into specific articles.
Auto-categorizing feeds
When feeds lack categories (e.g. after a bulk merge):
rss --json list to get all feeds with their titles/URLs- Analyze the feed titles and URLs to infer topics (e.g. "simonwillison.net" -> Tech/AI, "krebsonsecurity.com" -> Security)
rss recategorize <query> --category <CATEGORY> for each feed
This leverages the agent's own reasoning rather than hardcoding classification rules.
Curated briefing
Combine search + digest for a focused briefing:
rss digest --since 7d --limit 100
rss search "LLM" --limit 10
rss search "rust" --limit 10
Synthesize findings into a summary for the user.
Tips
- Use
--json for piping to jq or further processing.
- Cache is per-feed URL, stored under
~/.cache/rss-reader/. Use --refresh to force re-fetch.
- The OPML file is standard format -- importable/exportable with any RSS reader.
- Full
content:encoded is rendered when available; falls back to description summary.
search and digest fetch all feeds, which can be slow with many subscriptions. Use cached results when possible.
