google-ads-builder
暂无简介
- License 许可
- MIT
- 最近更新
- 2026-11-10
- Stars
- ★ 13
- Forks
- 4
02 资源文件 5
- 其他.gitignore111 B
- 参考文档.hivemind-skill-source.json2.9 KB
- 许可证LICENSE1.1 KB
- 参考文档README.md2.4 KB
- 脚本scripts/render_report.py12.0 KB
03 SKILL.md 正文
Google Ads Builder
The problem this solves: spinning up a Google Search campaign from scratch is hours of tedium — keyword research, grouping into tight ad groups, writing 15 headlines and 4 descriptions per ad group inside Google's exact character limits, building a negative list so you don't burn budget on junk clicks, and wiring up extensions. Most people either pay an agency for the setup or let Google's "Smart" defaults build a sloppy campaign that leaks budget.
This skill does the whole build from one input: your website URL. It reads the site, works out what you sell and who for, and produces a complete, launch-ready Search campaign — structured the way a good paid-search manager would build it — exported as a Google Ads Editor CSV you import in one click, plus a visual dashboard.
What it is NOT: it cannot create or activate a live campaign, connect to your Google Ads account, or access account credentials. It builds the campaign draft; you review it, add conversion tracking + real budgets, and launch. Search-volume figures are directional estimates (no paid keyword-tool API) — validate the final keyword list in Google Keyword Planner before scaling the budget.
0 — Intake
Get the website URL (required). Then ask (don't skip — they change the whole build):
- What's the primary conversion? purchase / lead form / phone call / booking. (Shapes the CTAs and bidding.)
- Target geography? country / region / city radius. (Default: the site's apparent market.)
- Rough monthly budget? (Shapes how many ad groups + match-type aggressiveness. Optional.)
- Any competitor brands to target, or brand terms to protect? (Enables competitor + brand ad groups. Optional.)
Save intake to ./search-ads-run/brief.json.
1 — Read the site
Fetch the homepage plus the key pages (products/services, pricing, about) with your web tools. Extract and note in brief.json:
- What they sell (products/services, categories)
- Core value props + differentiators (free shipping, guarantee, made in X, awards, speed)
- Price points / offers (so ad copy can lead with a real hook)
- Proof (reviews, customer count, notable clients)
- Geographies + audience signals
If the site is JS-heavy and a raw fetch returns an empty shell, use a headless/browser web tool. If you can't read the site at all, stop and say so — a campaign built from a guess isn't worth shipping.
2 — Design the campaign structure
Reason about the account before writing copy. Build tightly-themed ad groups — never one giant bucket. Standard shape:
- Category / high-intent — bottom-funnel buyers (
buy <product>,<product> for <use>,best <category>). One ad group per distinct product line or theme. - Problem-aware — searchers describing the pain your product solves (
how to <problem>,<problem> solution). - Brand — your own brand terms (cheap, high-converting; defends against competitors bidding on you).
- Competitor (optional, only if provided/derivable) — competitor brand terms. Flag the higher CPC + policy limits (no using competitor brand in ad text).
Each ad group: 5–15 tightly related keywords, mostly phrase and exact match (broad only with Smart Bidding + tight negatives). Note match type per keyword.
3 — Negative keywords
Build a negative list so budget isn't wasted:
- Universal junk:
free,jobs,salary,diy,how to make,used,torrent,pdf(tune to the business). - Business-specific: irrelevant meanings, wrong audience, out-of-scope products.
- Match-level: which negatives are exact vs phrase.
4 — Write the Responsive Search Ads
One RSA per ad group. Follow Google's spec exactly:
- 15 headlines, each ≤ 30 characters. Mix: keyword-matching, benefit, offer/price, CTA, proof, differentiator. Pin the brand or primary keyword to Position 1 only where it must always show.
- 4 descriptions, each ≤ 90 characters. Lead with the benefit/offer; end with a CTA.
- Policy-safe: no excessive capitalization, no unsupported superlatives ("#1", "best" without proof), no phone numbers in headlines, no repeated punctuation. Every claim must be supportable from the site.
- Include the ad group's core keyword in ≥ 3 headlines for Quality Score.
Character counts are non-negotiable — the export/render step flags any overflow, but write them correct the first time.
5 — Extensions + settings
- Sitelinks (4): real pages from the site with a 25-char text + two 35-char descriptions each.
- Callouts (4–6): 25-char benefit snippets (
Free Shipping,30-Day Returns). - Structured snippets: a header (e.g. Types, Brands) + values.
- Settings recommendation: match-type strategy, bidding (Maximize Conversions to start, tCPA once you have data), budget split across ad groups, geo, ad timing, network (Search only — Display off).
6 — Write campaign.json
Write the full structured campaign to ./search-ads-run/campaign.json:
{
"business": "name",
"url": "https://...",
"goal": "purchase|lead|call|booking",
"geo": "...",
"budget_monthly": 3000,
"settings": { "bidding": "...", "networks": "Search only", "timing": "...", "notes": "..." },
"ad_groups": [{
"name": "Category — <theme>",
"theme": "high-intent|problem-aware|brand|competitor",
"keywords": [{ "text": "buy blue widgets", "match": "phrase" }],
"rsa": { "headlines": ["...15, each <=30..."], "descriptions": ["...4, each <=90..."], "paths": ["path1", "path2"] }
}],
"negatives": [{ "text": "free", "match": "phrase" }],
"extensions": {
"sitelinks": [{ "text": "...", "desc1": "...", "desc2": "...", "url": "..." }],
"callouts": ["Free Shipping", "..."],
"snippets": { "header": "Types", "values": ["...", "..."] }
},
"caveats": ["honest limits of this build"]
}
caveats is mandatory. Always include: "This is a launch-ready draft, not a live campaign — add conversion tracking and set real budgets before launch. Search-volume and CPC are directional; validate in Keyword Planner." Plus run-specific limits (site partially readable, no competitor data, etc.).
7 — Render the dashboard + export the CSV
python3 ${CLAUDE_SKILL_DIR}/scripts/render_report.py
(If CLAUDE_SKILL_DIR isn't set, use the folder containing this SKILL.md.)
Reads campaign.json, writes to ./search-ads-run/:
google-ads-editor.csv— import-ready (Campaign / Ad Group / Keyword / Match Type rows + RSA rows + negatives). Import via Google Ads Editor → Account → Import.campaign-dashboard.html— dark-mode visual: structure tree, per-ad-group keyword + RSA preview (with live character counts, red if over limit), the negative list, extensions, and the settings recommendation. Opens automatically (--no-opento skip).
The script only renders + exports; the campaign design is model judgment.
8 — Report in chat
Close with: the campaign at a glance (N ad groups, M keywords, the budget split); the single sharpest keyword theme (where the cheapest high-intent conversions are); the one thing to fix before launch (usually: install conversion tracking); and the one-sentence takeaway — "a paid-search manager's first-week build, from just your homepage URL." Tell them the CSV path and that it imports straight into Google Ads Editor.
Notes
- Zero account credentials, no Google Ads login. The site is public and read with your built-in web tools; the renderer/exporter is stdlib Python.
- Tight ad groups beat big ones. Quality Score, CTR, and CPC all reward keyword↔ad relevance. Never dump every keyword in one group.
- The build is the product. The script only renders + exports; the keyword strategy, grouping, and ad copy are model judgment — written to Google's real character limits and policy every time.
- Tone of the output: a sharp paid-search manager handing over a launch-ready account, not a keyword dump. Every ad group has a reason to exist; every headline earns its place.
04 同仓库更多 Skills
tailscale-generation-skill-authoring
This skill should be used when the user asks to "create a Tailscale generation skill", "add a local generation skill", "document a generation endpoint", "add an image generation provider", "add a TTS provider", "add video generation", "add 3D generation", or needs guidance on where HivemindOS Tailscale or local generation providers belong in the shared skills shelf.
localtts
This skill should be used when the user asks to "generate local TTS", "use local text to speech", "use remote TTS over Tailscale", "make speech audio", "add a TTS provider", "use Qwen3-TTS", or needs TTS endpoint setup through local or HivemindOS Tailscale routes.
karpathy-guidelines
Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria.
comfyui-image-generation
This skill should be used when the user asks to "generate images", "generate images with ComfyUI", "use local ComfyUI", "use remote ComfyUI", "use Z-Image", "add an image generation provider", or needs image-generation endpoint setup through local or HivemindOS Tailscale routes.
wrangler
Cloudflare Workers CLI for deploying, developing, and managing Workers, KV, R2, D1, Vectorize, Hyperdrive, Workers AI, Containers, Queues, Workflows, Pipelines, and Secrets Store. Load before running wrangler commands to ensure correct syntax and best practices. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.
workers-best-practices
Reviews and authors Cloudflare Workers code against production best practices. Load when writing new Workers, reviewing Worker code, configuring wrangler.jsonc, or checking for common Workers anti-patterns (streaming, floating promises, global state, secrets, bindings, observability). Biases towards retrieval from Cloudflare docs over pre-trained knowledge.