New Term Radar Agent Skill

New-term radar

The whole job in one run: collect what is new, then work out which keywords are worth acting on. If the user wants only the collection, or only the keyword research on names they already have, the split skills ai-listening and seo-new-keywords-research do each half.

Before you run anything

• Find the radar: $NEW_TERM_RADAR_DIR if set, else ~/new-term-radar; clone https://github.com/zhuyansen/new-term-radar there if missing. Run every command from that directory. Python 3.10 or newer; nothing to install.
• An AIsa key must be available, as AISA_API_KEY or from the AIsa CLI (aisa login stores it in ~/.aisa): the X posts, the search volumes, Google Trends and the models all go through it. Without it, stop and say so. A key comes from https://go.aisa.one/yansen-zhu.
• GOANYAPI_API_KEY is optional (a key comes from https://goanyapi.com/r/ga-95708898fb). With it the table also holds what people really searched together with each name, and reads the bigger rows from one country's settled series. Without it, tell the user a "flat" reading is weaker evidence.
• If there is no config.json, copy config.example.json and tell the user they are on the example list of X accounts and sitemaps.

Choose the path

The user has names already ("research PhotoCraft and SynthID Detector"): skip collection.

python -m radar keywords "photocraft" "synthid detector" --repo "photocraft=storytold/photocraft" --json

About 0.05 USD a name plus 0.09 USD a call, 40 to 90 seconds a name, eight names at most. One call for all the names.

The user wants today's new terms: run everything.

python -m radar daily --out report.md

With the example config this is about 0.30 USD and 10 to 15 minutes: it reads every X account in config.json (0.003 USD and 10 seconds each), finds breakout repositories and new sitemap pages, scores which terms are names, then builds the keyword table for up to five of the day's names. It writes report.md and report.json.

Say the cost and the time before either command, and do not run daily twice in a row: the second run reports only what changed since the first.

The first daily run on a machine only builds snapshots, so it reports GitHub breakouts and little else. Say so; do not present it as a quiet day.

Read the result honestly

From report.json: leads are the day's names (with kinds, signals, named, examples, posts), keywords is the keyword table (rows, tails, cosearch, and the error fields). From keywords --json the table is the whole output.

Each table row: phrase, kind (name or generic), names (the name it sits behind), new, per_day, started, peak_day, momentum, now, decision, brand, source, trend_geo, evidence.

• per_day is an estimate, good to an order of magnitude.
• now is rising, holding, fading or flat. A row with trend_geo was read from one country's daily data settled to yesterday. A row without it was read from a worldwide series whose last two days are discarded, where a real move can read flat: for those rows say "no rise seen", never "did not rise".
• brand true (⚠): the keyword carries someone else's brand. Keep it, flag it; using it is the user's call.
• source "searched" (🔎): people searched it together with the name. Stronger than a phrase the model proposed.
• If posts_error, related_errors or country_errors is set, or a signal's block in report.json holds error, a source failed. Say which, and that the result holds less than it should. A thin table is not a quiet market.
• If keywords.names is empty, no lead today was a thing with a name score of 0.8 or more. Say that; offer to run keywords on a name the user picks from the leads.
• Names usually rise within days of release and may fade within a week. Generic terms move later and less, if at all.

Deliver

• Today's new names (skip this part on the names-given path): strongest first, each with what it is in a few words from the examples, where it was seen with the number that shows its strength, and one link.
• The priority table, sorted prioritize, watch, skip:

| Keyword | Type | Behind | New? | Est. searches/day | Started rising | Momentum (7d) | Now | Decision | Why |

Use the radar's decision and add one short reason per row from the fields above.

• Under the table: each name's long-tails (tails), other names found alongside (cosearch), then every caveat that applies: missing key, failed source, first run, estimate quality.
• Two or three lines of advice: which rows are worth a page or a tool now, and what to look at again in a few days. Search volume data runs about two months behind; say so if the user is about to commit work to a number.

Do not add names, rows, volumes or dates from your own knowledge. The output is what the radar saw.