Install any skill in seconds. Free to start, no credit card required.
Get Started Free →justinbao19/seo-blog-writer
Fully automated SEO article writer. Give it a topic and domain — it handles everything: auto-discovers product context, researches keywords, analyzes competitors, writes the article, verifies links, and delivers a complete publication package (article + QA report + schema markup + promotion checklist). One confirmation gate before writing starts, then hands-off execution. Works for any site, any industry, any CMS. No hardcoded dependencies.
Passed 8 of 18 test cases with the skill, and 9 of 18 without it.
and 3 more in the full report
4 fixed4 passed either way5 still failing5 broke4 errored
Tokens, turns and time describe resource use, not a dollar price.
One of 4 test cases the skill fixed. Same model, same prompt, with and without the skill.
Please adapt our English guide on database indexing strategies from cloud-tech.example.com for a Chinese developer audience. Rebuild the syntax and technical examples so they read naturally in Chinese while preserving all factual evidence, and save the final localized article draft with checked image references in the workspace.
| Checked | Without skill | With skill |
|---|---|---|
| The localized article draft is saved inside a slug directory under seo-output. | Missed | Passed |
| All image references in the localized draft use the .webp file format. | Missed | Passed |
The instructions your model reads when the skill loads.
Turn a topic and target website into a fact-supported article or directory collection that answers a clear search intent. Use existing product context and the requested mode; infer routine editorial choices. Complete research, drafting, correction, and the requested local package without an automatic discovery/outline approval gate.
Write for a particular reader, publication, and language. Let the reader's question and the available evidence determine the article's shape; do not force keyword placement, word counts, FAQ sections, or a generic voice. For substantial English writing, study the target site's published work and, when useful, a few genre-matched editorial examples. Learn their editorial moves, not their sentences. For Chinese, write in natural Chinese from the underlying facts and reader need; when localizing an English draft, preserve meaning and evidence while rebuilding syntax and examples for the Chinese reader. Read voice editing before any full English or Chinese article or style revision.
For topic or brief-only requests, deliver that output. For full articles, continue to the final artifact and relevant QA. Ask only about consequential ambiguity, missing access, or an explicit review checkpoint. Use tools actually available in this session; do not assume an external agent runtime or change permissions to follow an example.
Keep factual evidence traceable. Do not invent keyword volumes, source quotations, competitor features, measured rankings, or traffic forecasts. Mark estimates and unresolved evidence. Blocked search or optional SERP tooling can reduce coverage without blocking a useful, accurately scoped result.
Save substantial research and draft milestones for resumption. After interruption, inspect existing artifacts and continue from what remains. Validate the finished artifact and changed areas; repair concrete failures, stopping after required checks pass. Do not run mechanical write/read cycles after every paragraph.
Choose references by the work needed; do not load all phases at once:
The existing blog format requires embedded images and cover assets to be genuine WebP files with valid references and descriptive text; preserve original source/provenance. Run that check when images or their references change and at final image QA. Respect a user-specified destination's different format contract.
Use the user/environment output directory when supplied; otherwise use seo-output/<slug>/ in the task workspace. Drafting and packaging alone do not authorize publishing, external uploads, or recurring promotion. Report the actual validation and remaining limitations, without promising rankings or claiming a local package is published.
test cases passed, vs 9 / 18 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.