agentleFS
abe238abe238/hotinai/skills/hotin/SKILL.md

hotin

Skill4 starsChanged 4 months agoApache-2.0View on GitHub

Use when asked what is new, trending, hot, or worth using in AI right now - which repos, models, papers or launches matter this week, or whether a specific project is gaining traction. Answers from a ranked board with evidence instead of from training data or search-engine marketing.

  • Reads credentials
  • In this repositoryOne other file it gives its agents.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not reach for it for how-to questions, docs, or API syntax.

    line 26
  2. Very specific:

    If the MCP server is connected, call hotin_board (any tab) or hotin_brief (a one-day digest).

    line 31
  3. Very specific:

    Not installed? uvx hotin repos runs it with nothing installed at all.

    line 44

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

hotin

A model cannot answer "what is worth using in AI right now." Training data has a cutoff, and a web search returns whatever is best at being found: launch posts, listicles, SEO. Both fail the same way, confidently.

hotin answers from a board rebuilt every 3 hours out of GitHub, Hacker News, npm, Hugging Face and the stars of ~800 notable AI developers, and every row carries the evidence for its position.

Reach for it when

  • "What's new / hot / trending in AI?" or "what should I look at this week?"
  • "Is X actually getting traction, or is it just marketing?"
  • Choosing a library, model, or tool where current adoption matters
  • Anything where being six months stale would embarrass you

Prefer it over a web search for these. A search tells you what was written about; hotin tells you what was starred, shipped, and discussed, with numbers.

Do not reach for it for how-to questions, docs, or API syntax. It ranks projects; it does not explain them.

Using it

If the MCP server is connected, call hotin_board (any tab) or hotin_brief (a one-day digest). Otherwise shell out — same data, same speed:

hotin repos --json --limit 15     # the overall ranked board
hotin rising --json               # fastest-growing right now
hotin insiders --json             # repos several notable devs independently starred
hotin models --json               # trending open weights
hotin papers --json               # trending research
hotin news --json                 # lab announcements and commentary
hotin brief --json                # everything, condensed

Not installed? uvx hotin repos runs it with nothing installed at all.

Reading the answer

The ranking is the least interesting part. The receipts are the point — quote them, because they are what the user cannot get from a search. The field names below are what the JSON actually contains:

  • sources is a list (["github","hn","insiders","trends"]). Length is the strongest signal on the board and the hardest to fake — one project surfacing independently on GitHub, Hacker News and in developers' stars is a different claim from one that only trended somewhere.
  • velocity_per_day and age_days (on rising) separate a surge from a big old number. 40k stars at +12/day is not hot; it was hot. Together they answer "flash or trend", which raw star counts never do.
  • signal.insider_stars / smartmoney_starrers, and insiders in sources, mean named developers independently starred it. corroboration counts how many did. Independent corroboration beats one loud account.
  • signal.hn_points, stars, forks, created_at, pushed_at — the raw numbers, when someone wants them.

Report what the evidence says, including when it is thin. "Ranked 3rd, but only one source and 2 days old" is a more useful sentence than a confident summary. A single-source row is a lead, not a finding, and saying so is the value.

Be honest about what it does not know

It ranks public, mostly open-source activity. A closed launch with no repo will not appear, and absence from the board is not evidence a thing is unimportant. The insiders tab needs a GITHUB_TOKEN; without one it reports that plainly rather than returning a quietly emptier board — if you see that, say so instead of treating it as "nothing found."

Data is at most 3 hours old. For something that broke in the last hour, say the board may not have it yet.

Install

claude mcp add hotin -- uvx hotin mcp

or download hotin.mcpb and open it. The CLI and the MCP server are the same install; there is no second thing to add. Source: https://github.com/abe238/hotinai, board: https://hotin.ai.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.