← ClaudeAtlas

trackiq-amazon-search-term-harvesterlisted

Finds the Amazon search terms that already convert above break-even but have no keyword of their own, and turns them into an upload-ready Sponsored Products bulk file with exact-match keywords and suggested bids. Checks every candidate against the existing targets so nothing is proposed twice. Use when the user asks about harvesting search terms, promoting search terms to keywords, mining auto campaigns, converting search terms, keyword expansion, what should we add as exact match, or finding new keywords from the search term report.
TrackIQ-HQ/trackiq-amazon-search-term-harvester · ★ 0 · Data & Documents · score 72
Install: claude install-skill TrackIQ-HQ/trackiq-amazon-search-term-harvester
# Search-Term Harvester The mirror of the Wasted-Spend Sweeper. That one cuts what is not paying; this one **promotes what already is.** A search term converting above break-even inside an auto or broad campaign is being bought at whatever bid the match type happens to produce. Given its own exact-match keyword it can be bid deliberately. Output is a branded HTML report plus an Amazon Sponsored Products bulk file. ## Requires - The TrackIQ MCP, for `list_marketplaces`, `get_search_terms`, `get_targets` and `get_ad_groups`. - **A break-even ROAS** — ask for it. Gross margin divided into 1. Without it the report cannot say which terms are worth promoting, only which convert. Default 2.0x, labelled an assumption. - **A destination ad group.** The search-term data carries no campaign or ad group, so nothing can be routed automatically. See non-negotiable 1. - Nothing else. No filesystem, no shell, no internet. `assets/build_bulk.py` writes the workbook if a shell is available; the format is in `assets/bulk-format.md` either way. - **Without the MCP:** works from a Search Term Report export plus a keyword export for the dedupe. ## First run Fill in a copy of `assets/account.example.md` saved as account.md beside the skill. Every TrackIQ skill reads the same file, so an account already set up for another TrackIQ report needs nothing added here. If the runtime has no filesystem, print the same block and ask the user to paste it into their project instructions o