What does this skill do, and when should you use it?
This skill provides a recurring workflow for mining paid-ad search-term reports and maintaining negative-keyword hygiene. It harvests converting queries into keyword or ad-group candidates, then identifies queries and n-grams that spend without converting. Its outputs include an add/negate/move maintenance diff, a standing negative-keyword list, an n-gram waste report, and the ROAS S spend-efficiency score. It does not handle account structure, budget or bid allocation, or the final RQS.
Reads search-term exports containing query, impressions, clicks, cost, conversions, and optionally conversion value; confirms a direct-response, prospecting, or incremental-profit profile; filters converting queries above a stated threshold and routes them to new keywords or matching intent ad groups; flags non-converting queries above a stated cost floor and assigns exact or phrase negative match types at ad-group, campaign, or shared-list level; tokenizes non-converting queries into 1-, 2-, and 3-grams and ranks tokens by measured cost; produces add, negate, and move rows, a standing negative list, an n-gram waste report, an S score, and a handoff summary.
- A paid-search team receives a weekly search-term export and needs to promote converting queries into the existing keyword or ad-group structure.
- An advertiser sees meaningful spend with zero conversions and needs a documented negative-keyword list with match types and levels.
- A growth team wants to find recurring tokens that drain spend without producing conversions.
- A paid-media lead needs search-term waste evidence and an ROAS S input before an account audit.
How do you install this skill?
- Before applying negatives or moving queries between ad groups, obtain user confirmation of thresholds, match types, scope levels, and each recommendation to avoid suppressing valuable traffic.
- Search-term exports may contain customer queries, costs, and conversion information; the skill does not define redaction, data minimization, or memory cleanup procedures.
- Confirm the profile, conversion column, attribution window, and platform-field semantics before scoring S or classifying waste.
- Tokenization rules for Chinese queries are unspecified; applying English-style n-grams directly may produce misleading findings.
- Network access
- Local filesystem
- MCP Server
Install the repository in Claude Code or another Agent Skills-compatible host, then use ad/research/search-term-miner/SKILL.md. The README documents these collection-level commands: /plugin marketplace add aaron-he-zhu/aaron-marketing-skills, followed by /plugin install aaron-marketing@aaron for Claude Code, or npx skills add aaron-he-zhu/aaron-marketing-skills for generic hosts. The README shows a single-skill example using keyword-research, but does not document the exact single-skill command for this skill.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/aaron-he-zhu/aaron-marketing-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/ad/research/search-term-miner" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Provide a search-term report with query, cost, and conversions columns, adding conversion value when ROAS scoring requires it. Example prompts: Mine my search terms. Here is my exported search-terms report: [paste/path]. Goal is [DR/prospecting]. or Build a negative-keyword list and an n-gram waste report from this search-terms export: [path]. If the conversion column is missing, the skill stops and asks for it rather than mining or negating from clicks alone.
What are this skill's strengths and limitations?
- Defines concrete add, negate, and move maintenance outputs.
- Combines query-level negative hygiene with a cost-ranked n-gram waste report.
- Requires figures to be labeled Measured, User-provided, or Estimated.
- Works from user-provided exports and does not require a paid-ad API.
- Requires query, cost, and conversion data; it cannot proceed without the conversion column.
- Does not design account structure, allocate budget or bids, or calculate final RQS.
- Thresholds, cost floors, and negative levels must be stated for the run; the source does not define universal defaults.
- The source does not provide an independent test suite or complete validation across specific ad platforms.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Paid Search Term Miner this page | 54 · Use with care | ★ 3k | 13d ago | Apache-2.0 |
| Paid Ads Campaign Architect | 57 · Use with care | ★ 3k | 13d ago | Apache-2.0 |
| Ad Fatigue & Frequency Manager | 58 · Recommended | ★ 3k | 13d ago | Apache-2.0 |
| Paid Ads Audience Segment Builder | 56 · Use with care | ★ 3k | 13d ago | Apache-2.0 |
| Paid Measurement Loop | 54 · Use with care | ★ 3k | 13d ago | Apache-2.0 |
Compared with campaign-architect, this skill owns search-term harvesting and negative hygiene while that skill owns account structure. Compared with budget-optimizer, it addresses query-level waste rather than budget and bid allocation. Compared with ad-account-auditor, it supplies the ROAS S evidence while the auditor calculates the full RQS and applies the audit gate.
How did FollowSkills review this skill?
The skill treats exports as untrusted input, relies on user-provided manual exports by default, keeps APIs optional, and limits persistence to user-confirmed local memory; it does not clearly classify sensitive ad data, define deletion/rollback, or require item-level confirmation before acting on recommendations, so points are deducted.
Column validation, stop-on-missing-input behavior, threshold-based routing, match-type labeling, and user-facing blocking are reasonably clear; however, harvest thresholds, waste cost floors, and malformed-data handling are unspecified, and there are no skill-specific key-path tests, limiting the static score.
Triggers, inputs, outputs, and non-fit boundaries are clearly described, and manual CSV input reduces platform dependency; evidence for platform-field variation, Chinese query tokenization, and mainland-China environment fit is limited, with geo-relevance explicitly marked low.
The skill has version, license, compatibility, Quick Start examples, contract, data sources, scope guard, and next-skill guidance; it lacks representative output examples, FAQs, named maintenance responsibility, changelog detail, and an independent update path, while publisher identity is unverified.
The defined maintenance diff, harvested keywords/ad groups, negative list, n-gram report, and S score are directly aligned with the core mining task; however, thresholds and cost floors require judgment, and no representative output or real-result evidence is supplied, so the static ceiling applies.
The document specifies fields, method, and Measured/User-provided/Estimated labeling; it contains no reproducible skill-specific tests, execution records, or third-party corroboration. Repository tests cover architecture and audit artifacts rather than this skill's key paths.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →