DataForSEO Skill — keyword & SERP research

Purpose: Get real search demand (monthly volume, competition, CPC) and keyword ideas from Google Ads data via DataForSEO, so naming and content decisions are grounded in data rather than intuition.

Auth (never commit the key)

DataForSEO uses HTTP Basic auth. Provide credentials via environment only:

export DATAFORSEO_LOGIN="you@example.com"
export DATAFORSEO_PASSWORD="…"
# or a single combined value:
export DATAFORSEO_API_KEY="login:password"   # or its base64

The helper reads these from the env — it never hardcodes or logs the secret, and only ever calls api.dataforseo.com. Keep the key in a local .env that is git-ignored; do not paste it into the repo, a URL, or chat.

Usage

The helper is dataforseo.py in this skill folder:

DFS="$(cat ~/.claude/.bootstrap-dir)/skills/dataforseo/dataforseo.py"

python3 "$DFS" whoami                                  # auth check + balance
python3 "$DFS" volume "claude code" "coding agent"     # search volume table
python3 "$DFS" ideas  "coding agent"                   # related keyword ideas
# options: --location <code, default 2840=US>  --language <code, default en>

volume → each keyword's monthly search volume, competition (LOW/MEDIUM/HIGH), and CPC. ideas → the searched long-tail around a seed.

Reading the results (naming decisions)

• Zero / n/a volume on every candidate is itself the answer: the names aren't search terms, so SEO can't rank them against each other — decide on clarity/brand and accept that discovery happens on GitHub/HN/X, not search.
• Category term vs. brand term: a huge-volume category word (e.g. the tool's ecosystem name) is where demand lives; a brand name rides it only if it contains it. Weigh discoverability against being on-brand for what you built.
• Wrong-intent volume is a trap: a candidate with existing volume may be drawing an audience searching for something else (a different industry meaning). That volume hurts, it doesn't help.
• Competition + CPC proxy commercial intent, not just popularity — high CPC means advertisers pay for that term (buyers), low CPC often means informational or noise.

Cost & etiquette

live endpoints bill per call (search volume / ideas ≈ $0.05–0.10 each). Batch many keywords into ONE volume call (it accepts up to 1000) rather than looping. Check whoami for balance before large runs.

Common endpoints (beyond the helper)

All take the same Basic-auth header; POST a JSON array of task objects.