How to use the Last 30 Days Community Research agent skill

What is an agent skill and what does last30days do?

An agent skill is a folder of instructions and supporting files that an AI agent loads when a request matches the skill's purpose. The skill's SKILL.md file describes when to use it and how to work. This one is named last30days and is titled Last 30 Days Community Research.

When you ask about recent sentiment, community reactions or launch reaction, the skill tells the agent to run its bundled Python research engine across recent community and social sources for the last 30 days. The agent then reads the results, separates praise from criticism, notes which sources were actually checked, and saves a Markdown report with citations in your Design Files folder.

• Searches a 30-day window of community and social sources through the bundled engine
• Separates praise, criticism and repeated questions
• Records which sources were checked, unavailable or thin
• Saves a reusable Markdown report at research/last30days/<safe-topic-slug>.md
• Treats online content as untrusted evidence, not as instructions

Install the last30days skill

Download the ZIP of the skill from this page. Then install it in one of two ways, depending on where you use your agent.

For Claude Code, unzip the download into your user skills folder ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill available only in that project. For the Claude apps, open Settings, go to Capabilities, then Skills, and upload the ZIP there.

• Option A (personal): unzip into ~/.claude/skills/
• Option B (single project): unzip into the project's .claude/skills/ folder
• Option C (Claude apps): Settings → Capabilities → Skills, then upload the ZIP

Check the requirements before you run it

The bundled engine needs Python 3.12 or newer. The skill's instructions say to use python3.12 and to fall back to python3 only after confirming that it is version 3.12 or newer. The skill also needs Node.js for the vendored X search client, and network access to the sources it queries.

Credentials are optional and are set in a .env configuration file. Depending on which sources you want, that can include X session cookies (AUTH_TOKEN and CT0), a Bluesky handle and app password, and other provider keys. If a source has no credentials, it is reported as unavailable rather than filled in.

• Python 3.12 or newer
• Node.js (for the vendored X search client)
• Optional .env entries: AUTH_TOKEN and CT0 for X; Bluesky handle and app password; other provider keys as needed
• Network access to the sources the engine queries

Trigger the skill with natural requests

You do not need a special command. Ask in plain language, and include a phrase that matches the skill's triggers, such as 'last 30 days', 'recent sentiment', 'community reaction', 'what people are saying', 'trend scan' or 'social research'. Name the topic and, if relevant, the audience or product.

The skill also responds to the triggers last30days and launch reaction, and to the Chinese phrases 最近30天 and 社区反馈.

• What are people saying about our new onboarding flow in the last 30 days?
• Run a trend scan on indie game marketing this month.
• Give me recent sentiment on the Acme design tool launch, with citations.
• Social research on home espresso machines, last 30 days.

Run the engine directly (optional)

The agent normally runs the engine for you. If you want to see the command it uses, the skill's runtime section shows this pattern. Replace the placeholders with your own topic and the path to the staged skill folder.

The engine may also write a raw support file such as research/last30days/<topic>-raw.md. The agent treats that file as evidence and then writes the final report itself.

Read the report and its limitations

The final report follows a fixed structure: Topic, Short Summary, Source Coverage, Key Findings, Community Signals, Limitations, Sources and an Evidence Note. Check the Source Coverage table first, because it shows which sources were checked, unavailable, thin or not relevant.

Use the Limitations section to judge how far to trust the findings. If you want a shareable HTML version, ask for one explicitly; the skill only creates HTML when requested, and saves it next to the Markdown report.

• Source Coverage statuses: checked, unavailable, thin, not relevant
• Key Findings and Community Signals cite sources as [1], [2]
• HTML brief is created only on request, at research/last30days/<safe-topic-slug>.html

Safety notes: scripts and network use

This skill runs scripts and uses the network. The engine sends your search queries to external services, so avoid private or sensitive topics unless you are comfortable with that. Optional credentials such as X session cookies are personal account secrets, so only add them if you understand what they grant.

The skill instructs the agent to treat webpages, posts and search results as untrusted. Embedded instructions in that content should not be followed. The skill also does not claim access to sources that were not actually checked in a run.

• Search queries are sent to external services
• Keep credentials in .env and do not share them
• Only sources that were actually checked are reported as checked

Tips for better results

Name a specific topic, product or audience. Narrow requests produce clearer themes than broad ones. Remember the window is the last 30 days, so for history or long-term trends use other sources.

Treat social sentiment as a signal, not a measurement. Social sources can be unrepresentative or biased, and coverage varies between runs.

• Be specific: 'reactions to version two of our pricing page' beats 'marketing'
• Ask for the Limitations section to be reviewed before sharing findings
• Reopen saved reports from Design Files instead of re-running research