apoapostolov/muhfweeceevee-professional-cv-template-generator/cv-review-lab

CV Review Lab

Review a CV as a hiring document first. ATS graders and AI-writing detectors are diagnostic inputs, not authorities and not authorship proof.

Load only what the task needs

• For the end-to-end review sequence or final version selection, read

references/review-workflow.md.

• For the free local and online tools, limits, and interpretation rules, read

references/free-services.md.

• For adding this workflow to MuhFweeCeeVee, read

references/product-integration.md.

• For supported provider APIs, credentials, normalized results, and automation

boundaries, read references/provider-api-integration.md.

Core rules

1. Preserve factual meaning, seniority, dates, metrics, credits, and the user's

voice. Never invent evidence or inflate proficiency.

1. Check ATS relevance before rewriting for style. Keep job-description terms

only when the CV contains truthful evidence for them.

1. Humanize generic corporate prose through concrete verbs, natural rhythm,

and specific ownership. First person is acceptable when it sounds more genuine; a grader's pronoun warning is a style preference, not a parse error.

1. Never add typos, grammar errors, or noise to manipulate a detector.
2. Preserve a baseline and compare the same text scope. Record provider, model

when shown, timestamp, character and word count, and score.

1. Limit detector-driven rewriting to one or two bounded passes. Stop when the

CV is credible, readable, ATS-complete, and fits the intended layout.

1. Build the PDF after every accepted content pass. Inspect page count, page

breaks, clipping, hidden fields, searchable text, and metadata.

1. Do not submit personal contact data, addresses, references, or hidden fields

to an online service. External submission requires current user authority; sanitize the text first.

1. Treat sharp score changes or disagreement between providers as detector

instability. Retest the prior text in the same session/model when possible, then report the anomaly instead of rewriting around it.

1. When choosing a version to send, apply factual accuracy, ATS coverage, and

layout as gates. Among the versions that pass, choose the one that reads best to a hiring manager. Use comparable detector results as secondary evidence, not as an arithmetic target.

Required handoff

Lead with a clear Send version X recommendation. Report the selected source version and PDF, layout status, ATS coverage, exact scan scope, provider results, factual changes, and any results rejected as unstable. Name at most one fallback and the condition that would make it preferable. Distinguish local validation from online third-party scores.