quality-evaluation
Quality Evaluation Skill
Evaluate Allied Brass generated product content using a performance-oriented rubric. This skill measures whether content would make a shopper click — not whether it followed the rules.
When to Use
- After generating content via the pipeline, to validate quality before approval
- When reviewing existing
candidate_contentin thegenerated_contenttable - When comparing content versions (before/after prompt changes)
- When creating gold standard examples for the
prompt_templatestable - When the
quality-evaluationskill is referenced by other skills (e.g.,quality_rubric.yaml)
Why the Current Scoring Fails
The pipeline's self_score in prompts.py uses 6 criteria:
| Current Criterion | What It Measures | The Problem |
|---|---|---|
specificity |
Does it mention specific product details? | Listing specs ≠ compelling copy. "18-inch center-to-center" scores high but reads like a database export. |
More from bobby-andris/allied-feedops
shopify-conversion-content
Generate Shopify product page content that converts browsers into buyers. Covers HTML description structure, title strategy, meta descriptions, and on-site conversion copy for luxury bathroom hardware. NOT for Shopping feed ads — this is storefront content.
27wrapup
Automated end-of-session workflow (CLAUDE.md update, continuation prompt, push)
4finish-expertise
Deep expertise on Allied Brass finishes for content generation. Use when generating finish sentences, variant descriptions, finish comparisons, or any content that describes a specific finish. ALWAYS use this before writing finish-related copy.
4preflight
Validate task approach and requirements before execution
4data-first
Gather and verify real data before any analysis or planning
4repo-audit
Audit repository for unused files, organizational issues, and outdated content
4