News

**Quick judgment:** A well-designed, profile-driven news briefing skill with strong editorial and safety practices, best suited for users who want structured, personalized summaries rather than raw news feeds. Its design is strong, but real-world usefulness will depend on how reliably the underlying agent fetches current, accurate news. **What it does well** - Builds and maintains a detailed local profile (`~/news/memory.md`) of interests, proportions (e.g., “70% AI, 30% markets”), format (bullets vs narrative vs headlines-only), and timing (morning/evening/weekly/on-demand). - Always checks memory before generating a briefing, so output should consistently reflect stored preferences. - Enforces good information hygiene: lead with *what* happened, then *why* it matters; include when the news broke; cite named sources; and use multiple sources for controversial topics while calling out editorial leanings and disagreements. - Tracks engagement in `~/news/history.md` and uses it to suggest profile refinements, making personalization more accurate over time. - Keeps user preference and engagement data local (`~/news/`), which is good for privacy and data control. **Key risks & limitations** - **News accuracy & freshness:** The spec emphasizes “never fabricate” and “don’t present stale news as fresh,” but the page text doesn’t show the concrete mechanism for sourcing up-to-date news. Actual performance will depend on the underlying agent’s browsing/tools setup and discipline, so there is residual risk of hallucinated or outdated events. - **Dependency on correct memory setup:** It assumes the `~/news/` directory and markdown files (guided by `memory-template.md`) are correctly initialized and maintained. Misconfigurations or missing files may degrade personalization or cause inconsistent behavior. - **No permanent content store:** The skill explicitly avoids storing news content permanently. This improves privacy but limits long-term cross-article analysis or rich historical recall beyond what’s abstracted into history notes. - **Controversial-topic handling quality:** While the policy is solid (multi-source, bias disclosure), practical quality will depend on the range and reliability of sources the agent actually uses. **Recommended scenarios** - You want **short, predictable news briefings** aligned to specific, nuanced interests (e.g., “US monetary policy decisions + major AI model releases”) with control over format and timing. - You value **editorial transparency and safety**: clearly separated facts vs analysis, multiple sources on contested topics, and explicit source/bias disclosure. - You are **privacy-conscious** and prefer that your reading preferences and engagement history stay in local markdown files rather than being sent to external services. - You’re comfortable with a setup where the agent reads/writes simple files in `~/news/` and possibly combining this skill with related tools like `summarizer`, `scrape`, or `reading` for more advanced workflows.