Triage RSS Feeds with Papr: A Token-Efficient Agent Skill

Where this fits

Imagine you're a developer who tracks dozens of RSS feeds for research, product news, or community updates. Your agent needs to catch you up on unread items, search for a specific topic, or summarize a feed—without burning hundreds of tokens on verbose JSON. The papr-rss skill wraps the papr CLI, giving agents a lean, structured interface to your local Papr RSS database. It turns a messy feed backlog into a fast, command-driven workflow.

Why agents benefit

• Token-minimal output: Emits TOON format (≈40% cheaper than JSON) by default, with long bodies truncated unless you explicitly request --full.
• Self-orienting: Running papr with no arguments prints the unread dashboard and next-step commands, so the agent can act without reading a manual.
• Structured errors: Exit codes (0, 1, 2) and diagnostics on stderr let agents handle failures gracefully—retry, flag usage issues, or report no-ops.
• Complete feed lifecycle: Agents can list, search, summarize, mark read/star, subscribe, and refresh—all from one CLI.
• Local and private: Works against a local database, so your reading history never leaves your machine.

Practical scenarios

Morning digest with zero context The agent runs papr to see unread counts, picks the top 3 items, fetches --full for one, and summarizes them—all without you typing a single search query.

Deep search across your archive You ask "find anything about vector databases from last month." The agent runs papr search vector database and filters by date, returning only relevant titles and URLs.

Quick subscription triage You stumble on a new blog and say "subscribe to this feed." The agent runs papr add <url>, validates the feed, and confirms the subscription—then refreshes your dashboard.

Add it to your agent workflow

Install the skill into your agent's environment with:

npx skills add l0ng-ai/papr --skill papr-rss

Then invoke it naturally. For example, to catch up on unread:

papr

Or to search your subscriptions for a term:

papr search "RAG patterns"

The CLI handles the rest—your agent just parses the TOON output and acts.

Compare before adopting

Before adding this skill, compare it with alternatives like rss2json or manual curl + parsing. Key checks: quality signals (does the CLI handle malformed feeds, timeouts, or duplicate entries?), maintenance freshness (last commit, open issues), and workflow fit (does your agent already use Papr? If not, you'll need to set up the database first). Also consider whether a simpler sqlite query would suffice for your use case—but you'll lose the token efficiency and pre-built commands.

Why it is worth tracking

With 524 stars and an active author, papr shows real community momentum. The focus on token efficiency and agent-first design makes it a strong candidate for any RSS-heavy agent workflow. Track it if you're building personal assistants, research tools, or content-monitoring agents—and evaluate it against your own token budget and feed volume. For more context, see the skill page for updates and community notes.