Skip to content
Back to skills

Quality Check

ASecurity

Final gate before publishing. Checks the /voice-guardian and /web-content-optimizer grades and that the sensitivity scan ran on work content. Orchestrates, doesn't duplicate.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 12, 2026
ai-agentsrustgit

Security analysis

A100/100

Scanned October 12, 2026

npx -y skills add prPMDev/my-portfolio-kit --skill quality-check --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Quality Check?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Quality Check
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/prpmdev-quality-check/badge)](https://www.skillsdirectory.com/skills/prpmdev-quality-check)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: quality-check
description: Final gate before publishing. Checks the /voice-guardian and /web-content-optimizer grades and that the sensitivity scan ran on work content. Orchestrates, doesn't duplicate.
argument-hint: "[content or 'full-site']"
allowed-tools: Read, Grep, Glob
---

# Quality Check

Final gate before publishing. You are the **gate**, not the inspector. QC trusts specialists, doesn't redo them.

Run before publishing or committing significant changes, after `/website-expert` builds or changes the site, or whenever the user asks "ready to publish?". QC is re-runnable, not a one-time gate.

## Gate Logic

Gate on the grades from `/voice-guardian` (all content) and `/web-content-optimizer` (web pages; N/A for non-web content). Ask the user or check recent output.

| Grade | Decision |
|-------|----------|
| **A** / **B** | PASS |
| **C** / **D** | FAIL — fix, re-run that specialist, then come back |
| **Not run** (or scan not run) | INCOMPLETE — run it first. Takes precedence over FAIL. |

**Work content** also needs a yes/no: did the sensitivity scan run (`/story-adapter`'s built-in scan or `/anonymizer`)? Not a grade.

If a specialist makes changes, re-run that specialist to verify, then come back to QC.

## Output Format

```
## Quality Check — Gate Review

**Content:** [what's being checked]

| Check | Specialist | Result |
|-------|------------|--------|
| Voice | `/voice-guardian` | [A/B/C/D/Not run] |
| Web | `/web-content-optimizer` | [A/B/C/D/Not run/N/A] |
| Sensitivity scan (work content) | `/story-adapter` or `/anonymizer` | [Yes/No/N/A] |

**[PASS / FAIL / INCOMPLETE]**

[If PASS:] All checks passed. Content approved.
- "Build it" — `/website-expert` populates the site with approved content
- "Build + deploy" — `/website-expert` builds, then push to GitHub Pages
- "Export content" — get the approved content as files, build later

[If FAIL:] [Area]: Grade [X]. `/[specialist]` can address this.
- "Fix and re-check" — address issues, then re-run `/quality-check`
- "Ship anyway" — publish as-is; your call
- "Fix later" — note issues, move to build with known gaps

[If INCOMPLETE:] Run `/[specialist]` for [area] first.
- "Run them now" — then re-run `/quality-check`
- "Skip and build" — proceed to `/website-expert` with known gaps
```

## Quick Spot-Checks (Optional)

If the user wants a quick pass without the specialists:

- [ ] No placeholder text visible ("[Your Name]", "Lorem ipsum")
- [ ] No obvious broken links on main pages
- [ ] Content loads on mobile

Recommend full specialist reviews before a major publish.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.