SolidMITOriginal

QA Swarm

Launches 3-5 parallel QA subagents against a site, each with a distinct remit (SEO, monetisation, freshness, accuracy), synthesises findings.

Entry verified April 20, 2026

The short answer

Content QA is not one job — it's 5 jobs (accuracy, freshness, SEO, monetisation, link health). Running them serially is slow; running them in parallel is what makes Claude Code's agent model actually earn its keep.

When to use it

After a content batch lands. Monthly on evergreen pages. Before a major rebrand.

Setup

  1. 1

    Save as ~/.claude/commands/qa.md.

  2. 2

    Skill spawns N agents with narrow briefs. Collects outputs, synthesises into ranked findings.

  3. 3

    Invoke: /qa <scope>.

Example

You: /qa pages/claude-code-mcp-stack.js
Claude: 5 agents dispatched. Synthesis: 2 P1s (stale link, missing schema), 4 P2s. Full report below.

Source & attribution

Author
Bryan Collins
Licence
MIT
Type
Original

Original pattern published under MIT — attribution preserved by convention, not licence requirement.

Caveats

Costs 5x tokens per run. Budget accordingly.

Browse more skills

47 skills across 10 categories, all licence-verified.