what is last30flames fr?

last30flames is a skill for AI coding agents that researches what is genuinely new about any topic within a recent time window, such as the last 30 days. Instead of relying on a model's own outdated training knowledge, it gathers fresh information from live sources and hands that research to the agent, which then writes a short summary grounded in real, cited sources. It pulls from five sources at the same time: Firecrawl's web search, which scrapes full pages rather than short snippets, plus Hacker News, Lobste.rs, Bluesky, and GitHub, all through their public, official APIs. It deliberately avoids Reddit, X, TikTok, Instagram, or any platform that requires a login, so it only uses data those platforms make openly available. The tool itself contains no AI model. It only collects and prints the raw source material, and the actual summarizing is done by whichever coding agent is using the skill, so there is no extra AI service or cost involved in the gathering step. No API key is required to use it, since it works out of the box with reasonable limits. Adding a Firecrawl or GitHub API key later simply raises the rate limits and lets more requests run at once, without needing to reinstall anything. It installs into any agent tool that supports the skills format, either through the skills.sh installer or, in Claude Code, through the plugin marketplace. Beyond a basic search, the skill can handle topics with more than one meaning by first running a quick check and then narrowing the search, compare two topics side by side such as one tool against another, save research to a local folder so it can be reloaded later without re-scraping everything, and warn you if saved research has gone stale. It can also produce a shareable, self-contained HTML summary with clickable citation links that works offline and can be dropped into Slack, email, or a document. The underlying engine is a small TypeScript project built on the Bun runtime with no external runtime dependencies, small enough to read through and audit in a few minutes.