
Aether Listen: Find People Asking for Your Product
Aether Listen finds public posts where people describe the problem your product solves and scores them against intent phrases you agree on. It is for founders, indie makers and small marketing teams who answer community questions by hand. Your agent fetches recent posts from Hacker News, Reddit, StackExchange and RSS feeds you choose, scores them, lists the questions asked, and can hold reply drafts for your approval. It never posts anything. This is the listening half of Aether Engine, a paid skill by the same publisher, and it works on its own.
What you get
scripts/listen.py(network only): fetches recent public posts for the searches, subreddits, StackExchange sites and feeds you name. Reads no files, writes nothing, prints JSON.scripts/signals.py(local only, no network code): readsaether.jsonand the fetched posts, stores them in.aether/aether.dbin your project, scores them, groups the questions and keeps reply drafts until you decide. Your decisions train a small Naive Bayes model. It starts ranking only after 20 decisions with at least 5 approvals and 5 rejections.SKILL.mdandreferences/profile.md: the agent's instructions, including reply draft boundaries, and theaether.jsonformat.
Requirements and install
Python 3.8 or newer (tested on 3.13). Standard library only: no pip installs, no API keys, no accounts, no git. Unzip the download: you get one folder named aether-listen. Put that folder in your agent client's skills folder. It is written for agent clients that read SKILL.md and can run a shell command, such as Claude Code or Codex CLI.
How to use
- "Find people asking for a tool like mine and show me the best matches."
- "List the questions my market is asking, with links."
- "Draft replies to the top three posts and show them to me before I decide."
On first use the agent asks what you sell, proposes intent phrases and sources, and writes aether.json once you confirm.
Sample run
A real run for a fictional uptime monitor, on posts fetched on 12 Nov 2026 (script paths shortened, output trimmed).
$ python scripts/listen.py --hn "uptime monitoring" > posts.json
$ python scripts/signals.py ingest posts.json
{
"new_posts": 40,
"scored": 40,
"worth_reading": 19,
"worth_reading_note": "counts every stored post worth reading, not only the ones this run added",
"by_source": {
"hackernews": 40
},
"unreachable": []
}
$ python scripts/signals.py questions 15
{
"questions": [
{
"question": "What do you use to monitor cron jobs and queue workers?",
"asked": 1,
...
Output format
Each command prints JSON, which your agent shows in the chat as a Markdown table. top gives each post's id, source, link, title, first 700 characters, score, how it was scored (the phrases that matched, or the model's estimate once it is active) and date, or unknown when the source gave none. questions gives the question, how often it was asked and up to three links. queue gives each waiting draft with the post it answers. It creates only aether.json, posts.json, your reply text files and the local database. An error is one JSON line on standard error.
Network and data
Only listen.py uses the network. It sends read-only GET requests, with no login, key or cookie, to:
hn.algolia.com: your search wordswww.reddit.com: the subreddit namesapi.stackexchange.com: the site and tag names- the host of each RSS or Atom feed URL you add
It follows redirects from those hosts. Nothing from your files is sent. The posts it collects are public and are stored only in your project folder, as the poster wrote them: they include the poster's public username and anything else they published, which can be an email address. The skill calls no AI service itself. Your agent client sends the conversation, including command output, to the model provider you use.
What is automatic and what asks first
Automatic when you ask for a search: fetching public posts, scoring them and storing them locally. Your agent is instructed to ask before writing aether.json and before every approve, reject or skip. No script can post, send or log in anywhere: you copy an approved draft and post it yourself. A decision on a draft is final; you can save a new draft for the same post. A draft that mentions your product must say you are connected to it.
Limits
- The first ranking is phrase matching. In the sample run, several of the 19 posts marked worth reading were job-seeker and launch comments, not buyers. Add what you see as negative phrases and score again with
rescore. - Hacker News is searched by your words, up to 40 newest results per search. For Reddit, StackExchange and feeds it reads the newest posts (25 from the subreddit tested, 50 per StackExchange site) and your phrases pick the relevant ones, so a busy source covers only a short time.
- No date filter. The sample run reached back to July 2026.
- Reddit and StackExchange limit anonymous requests. A source that fails is named and skipped.
- It does not look up or extract contact details, and it contacts no one.
- Follow the rules of each site you read and post on.
FAQ
Does it post replies for me? No. It stores drafts and your decisions. You post them.
Where is my data? In your project folder: aether.json and .aether/aether.db. Delete them to remove everything.
How is Aether Engine different? It adds keyword planning, page publishing, site audits and a writing check. This skill only finds and scores demand.
Licence
MIT. Copyright (c) 2026 Abdullah Rathore. The full text is in LICENSE.txt and the full source is in the package.
外部 API
hn.algolia.com · www.reddit.com · api.stackexchange.com · RSS or Atom feed hosts you add


