Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →justinbao19/content-qa
Quality assurance agent for content review. Runs as a separate reviewer against defined checklists, outputs PASS or FAIL with actionable fixes. Designed to be spawned as a QA sub-agent in a produce → review → revise loop. Supports: SEO blog, ad copy, social post, email sequence, landing page. v2.0: Added SEO quality scoring, GEO extractability check, search intent verification, quote accuracy verification, AI writing pattern detection, proportional link standards.
Passed 21 of 22 test cases with the skill, and 16 of 22 without it.
and 3 more in the full report
5 fixed16 passed either way1 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
When the optional SEO quality package is located in a custom directory at /opt/custom-skills/seo-qa rather than beside the current skill, what environment variable must be configured to point to that directory?
| Checked | Without skill | With skill |
|---|---|---|
| Identifies SEO_QA_SKILL_DIR as the environment variable name | Missed | Passed |
The instructions your model reads when the skill loads.
Review the supplied artifact against its audience, purpose and actual destination requirements. A review-only request produces findings; revise when the user requests fixes. Do not invent flaws or rewrite merely to satisfy a preferred style.
Read the draft, brief and available product/brand facts. Check material claims, quotes in their original context, comparisons, author experience, language, logic and usable next steps. Verify volatile facts from current primary sources. An accessible URL does not prove its claim; a blocked request is unverified, not automatically a dead link.
Use review checklists for the requested format only. Suggested word counts, link totals, FAQ, keyword density, CTA and rhetorical patterns are heuristics. Apply them as requirements only when the destination or brief actually sets that contract. Natural Chinese and English may use different structures. A phrase is not evidence of AI authorship.
For substantial SEO drafts, use an installed seo-geo-qa if available. This is a conditional technical dependency; absence does not prevent an editorial review. Report technical coverage as unavailable when it was not run. The optional wrapper is:
bash <this-skill>/scripts/run_qa.sh draft.md --skip-serp
Set SEO_QA_SKILL_DIR to a separately installed QA directory if it is not alongside this skill. The wrapper does not install or publish anything.
Report the content type, scope, verdict, critical issues with locations and evidence, warnings, and checks not performed. PASS means no supported critical issue was found within the reviewed scope; FAIL means a material error or unmet explicit requirement remains. Unverified important claims must be verified, qualified or removed. Never call an automated PASS publication approval, a ranking guarantee or proof of human authorship.
Critical examples: false feature/price, misquoted source, unsupported central assertion, changed product behavior, missing required disclosure or broken required output. Style alternatives and optional metadata are warnings or notes. Scores may summarize an agreed rubric but cannot override material correctness or substitute for evidence.
For revisions, recheck changed passages and outstanding findings. After three unsuccessful review rounds, report unresolved items and the specific decision needed. Keep substantial reports next to the article or in the user’s output directory, never in the installed skill.
test cases passed, vs 16 / 22 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.