Topic research, comparisons and discovery

last30days

Turn Recent Discussion into a Research Brief

A Skills-capable assistant runs the research engine to collect recent posts, comments, video text and project activity from connected sources. It organizes the evidence into a traceable brief, with separate paths for known topics, discovery and searching past research.

By mvanhorn

Setup & first taskView on GitHub
Project demo

Project demo

View example source

Hosts listed by the author

  • Claude Code
  • Codex
  • Cursor
  • Gemini CLI
  • GitHub Copilot
  • OpenClaw
  • Claude web & desktop
GitHub stars
65,835
Version
3.31.0
Checked
Categories
Content & Knowledge

Useful when you want recent community evidence before making a decision

Narrow the question to one subject and use case, such as a tool’s behavior in an existing project. Read the original evidence behind the brief before choosing a small experiment.

Pick a task and get started

Install from a terminal into the agent you actually use, then invoke an example in chat. Follow the first-run setup for sources you need; you can skip browser access and paid services and begin with an available source set.

Choose the goal for this run

Choose a named project and a concrete use case. On a first run, decide which sources to configure, begin with those already available, and add accounts only as needed.

  1. Install in an Agent Skills host

    Run in a terminal and select the agent you use. -g installs for your user across projects; omit it for project scope. The author also provides a Claude Code plugin route. Choose one route to avoid duplicate entries.

    npx skills add mvanhorn/last30days-skill -g
    Read the author’s setup guide

    Check the runtime and sources before researching

    Confirm Python 3.12 or newer and have the assistant load the actual installed Skill directory. Use /last30days doctor for source status and repair guidance. Browser-session access and third-party signup are choices in onboarding; skipping them can still leave useful sources available.

    Read setup and source-repair guidance

  2. Give your AI this first task

    These editorial examples follow the author’s documentation. They are not captured research runs. Replace the subject and purpose and confirm which sources are actually available.

    /last30days Claude Code TypeScript workflows --search reddit,hn,github I maintain an existing TypeScript project and want to understand how developers have actually used Claude Code in the last 30 days. Organize the brief around workflows, recurring problems and disagreements. Link to original posts or repository records and identify evidence dates. Separate official documentation, individual experience and your interpretation; use older material only as background. End with three small experiments I could try on a branch. Do not describe an unavailable source as a source with no discussion.

  3. Check what you receive

    • A workflow-based research brief with original sources, recent discussion and grounded follow-up experiments.
    • The engine’s source statistics and actual saved-file paths, with unavailable, timed-out or partially retrieved sources described accurately.
    • Open the key links and check dates and project identity. Ensure conclusions follow from the evidence rather than post counts or engagement alone.
    Read the author’s notes for this task
More about this skill

Maintained by mvanhorn, last30days combines a Skill that directs the assistant with a Python research engine. The assistant resolves the subject, plans queries and interprets evidence. The engine collects available sources, enriches them with comments or transcripts, and organizes material using relevance, time and engagement signals. General web search supplements this work; a normal search answer is not evidence that the Skill ran.

Topic research investigates a known question. Comparison mode researches each candidate before organizing common dimensions. Discovery first finds directions worth investigating. These modes have different workflows and outputs. Their view of recent public discussion is a sample, not a survey of every user, and personal accounts remain personal accounts.

The default research window is the last 30 days, but backends and evidence types affect coverage. An official X recent-search path may cover roughly a week, while some research material and older saved briefs provide background only. Unconfigured sources, rate limits, authentication failures and clean zero-match results are different outcomes. Check both evidence dates and actual coverage.

Research files normally go to ~/Documents/Last30Days/ on your machine; the directory is configurable and saving can be disabled. Existing briefs support offline search and a local library index. Ongoing monitoring additionally needs storage settings and a scheduler. Local generation, scheduled execution and public publishing are separate actions.

Sources and verification

Sources checked

Checked against the README, current Skill, source configuration and relevant task references. We did not install services, read browser sessions or execute these research examples.

Author documentationOriginal skill

Continue working with the research

For academic questions, move into systematic evidence review. For long-term note organization, explore Obsidian’s file workflows.

Academic Research SkillsWorkflows for research, writing, record screening and simulated review that keep sources and decisions.Obsidian SkillsSix skills for Obsidian notes, Canvas, Bases, app commands, web extraction and template batches.dbskilldontbesilent’s business and content methods as Agent Skills: a dispatcher routes your real task.BloomA course folder that writes each next lesson from your answers and annotations.