skills/resplendent-resume-builder/references/quality-controls.md

# Quality controls and Resplendence Report Card

Read this reference before every final user-facing deliverable. It governs a resume, CV, audit, job analysis, application response, cover letter, outreach message, LinkedIn copy, interview pack, offer comparison, reference list, and portfolio case study.

## Final-review sequence

1. **Fact check:** Trace every employer, title, role location or work arrangement, date, credential, metric, client, tool, claim of responsibility, compensation figure, result, and quote to user-supplied or verified evidence. Remove unsupported claims; use a confirmation question or clearly marked placeholder instead.
2. **Purpose and target check:** Confirm that the artifact answers the requested role, audience, decision, or prompt. For targeted resumes and application audits, rerun the requirement-to-evidence coverage gate in [job application](job-application.md), verify where high-priority supported terms landed, and surface unresolved must-haves or possible knockout items. Use relevant terminology only where it is truthful; flag gaps rather than concealing them.
3. **Medium check:** For resumes/CVs, preserve conventional headings, clear chronology, readable order, selectable text, and a single-column ATS-safe default. Check structured-field completeness as well as appearance: role locations, canonical titles, dates, education status, and explicit supported skills must survive extraction; compound-title or form-required education fields that remain ambiguous must be named. For any downloadable paginated career document, also read and satisfy [document production](document-production.md): use semantic paragraph spacing rather than blank spacer lines, render and inspect every page, verify extracted text order, and correct accidental blank/mostly blank pages, forced gaps, or orphaned headings before delivery. For forms, meet the exact question and limit. For all other materials, make the next action and evidence easy to find.
4. **Privacy check:** Keep contact, reference, compensation, client, and confidential business details to the minimum needed. Do not insert third-party contact information without the user's instruction and consent context.
5. **Editorial-economy pass:** Apply [professional editorial method](editorial-method.md). Resolve unnecessary exact, semantic, metric, verb, cross-section, and cross-document repetition. Confirm that each section and sentence performs a useful function.
6. **Natural-language pass:** Quietly revise the prose using the rules below.
7. **Score and report:** Rate the revised deliverable with the Resplendence Report Card. A score measures the visible draft and evidence, not the candidate's worth or probability of getting hired.

## Natural-language pass

Use these AI-Slop-Mop-derived checks while preserving the user's voice and any valid formal structure:

- Put the useful point first. Cut announcement phrases that delay it.
- Prefer supplied concrete actors, actions, context, scope, and consequences to abstractions, management shorthand, vague praise, and empty intensifiers.
- Give related ideas a real relationship—cause, contrast, condition, or consequence—rather than stacking dramatic fragments.
- Use active voice when it makes responsibility clearer. Preserve passive voice when the actor is unknown, irrelevant, or deliberately de-emphasized.
- Remove canned transitions, template-like “not X but Y” contrasts, staged questions, redundant conclusions, and slogan endings when they do not add meaning.
- Let sentence length follow the content. Avoid three or more same-shaped sentences, artificial fragments, or default em-dash chaining.
- Preserve intentional user wording, technical terms, quotations, and brand names unless the user asks for a change.
- Never add a number, anecdote, quote, date, result, or experience merely to make the writing sound lived-in.

## Resplendence Report Card

After revising, score every final deliverable on each criterion from **0.0 to 10.0**, using the available evidence. Calculate the overall score as the weighted average and round to one decimal:

`overall = (truth × .25) + (target fit × .20) + (impact × .15) + (clarity × .10) + (ATS/format × .10) + (voice × .10) + (editorial economy × .05) + (completeness × .05)`

| Criterion | Weight | What the rating measures |
| --- | ---: | --- |
| Truth & evidence | 25% | Claims, metrics, credentials, dates, and outcomes are supported; uncertainties are explicit. |
| Target fit | 20% | The draft answers the role, audience, or goal without unsupported keyword stuffing. |
| Impact & specificity | 15% | Concrete actions, context, scope, and outcomes are visible where evidence permits. |
| Clarity & scanability | 10% | A reader can find the value and evidence quickly in the chosen medium. |
| ATS & format integrity | 10% | Always assign a numeric score. For ATS-relevant artifacts, rate parsability, conventional structure, accessibility, and inspected page economy. When ATS is not applicable, rate format integrity instead: medium fit, accessibility, rendered-page quality when paginated, text order, and compliance with the requested constraints; note that the ATS component was inapplicable without changing the formula. |
| Voice & naturalness | 10% | The prose is direct, professional, non-formulaic, and faithful to the candidate's supplied voice. |
| Editorial economy | 5% | Every section and sentence earns its space; repetition is consolidated or intentionally differentiated without losing unique evidence. |
| Completeness & consistency | 5% | Required sections are present and related materials agree. |

Do not inflate the score to reassure the user. If evidence is thin, rate the supported draft and state the missing evidence. If an issue can be fixed from supplied information, revise first and score the revised version. A high score never cancels a material warning, missing fact, or trade-off.

Append this report to every final deliverable. Give one concise, deliverable-specific reason for every numeric rating. Each reason must point to visible evidence, a visible limitation, or a named unresolved fact in the actual output; repeating the generic rubric definition is not a reason.

```markdown
### Resplendence Report Card

**Overall:** X.X/10.0

| Criterion | Score | Reason |
| --- | ---: | --- |
| Truth & evidence | X.X | [What is verified and any material uncertainty.] |
| Target fit | X.X | [Which target needs are covered and which remain open.] |
| Impact & specificity | X.X | [What concrete action, context, scale, or outcome supports the score.] |
| Clarity & scanability | X.X | [What makes the value easy or difficult to find.] |
| ATS & format integrity | X.X | [What supports parsability/format quality and any observed limitation.] |
| Voice & naturalness | X.X | [What makes the language specific and human, or where it remains generic.] |
| Editorial economy | X.X | [What was consolidated and any repetition or low-value material that remains.] |
| Completeness & consistency | X.X | [What is complete and any missing or inconsistent item.] |

**Strength:** [the strongest supported quality]
**Next improvement:** [the single highest-value fact, revision, or decision]
```

For a simple question that does not produce a draft or audit, state that no deliverable was generated rather than manufacturing a report card. Otherwise, the report card is required.