agentsclimarketplace

Supplement scorer

Skill liaoliang1981/dietarysupplement-scorer/claude-skill/supplement-scorer

Deterministic dietary-supplement scoring engine — one shared core, with Claude Skill, ChatGPT, Gemini, and web calling layers. Scores formula, quality, and safety.

Install
npx -y skills add liaoliang1981/dietarysupplement-scorer --skill supplement-scorer

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Score, rate, evaluate, or compare any dietary supplement on three dimensions (formula, quality, safety) using a deterministic, evidence-anchored engine rather than the model's own guess. Brand-agnostic, works across markets and languages. Make sure to use this skill whenever the user pastes a supplement ingredient list, describes a vitamin / mineral / multivitamin / sleep / sports / herbal product, shares or photographs a supplement facts label, or asks things like 'is this supplement any good', 'which one should I buy', 'rate this formula', or 'compare these two' — even if they never say the word 'score'. The skill runs a Python engine for consistent, auditable numbers and supplies structure-function compliance guardrails.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.3 KB, as published. Nobody here has run it

Supplement scorer

Turns a supplement product into auditable scores on formula, quality, and safety (0-100 each, plus an overall). The numbers come from a deterministic engine, never from your own estimation — so the same product always gets the same score, and every number is explainable.

Workflow

1. Build the product into this schema

Gather the product from whatever the user gave you — a pasted ingredient list, a described product, or a supplement-facts photo (read the label; estimate doses you can't read cleanly and say so).

{
  "name": "Gentle Iron + Folate",
  "brand": "Acme",
  "certifications": ["GMP", "third_party_tested"],
  "ingredients": [
    {"name": "Iron", "nutrient": "iron", "form": "ferrous bisglycinate", "amount": 18, "unit": "mg"}
  ]
}

nutrient must be a key from references/ingredients.json (iron, folate, vitamin_d, vitamin_b12, vitamin_b6, vitamin_c, vitamin_e, vitamin_a, vitamin_k, calcium, magnesium, zinc, iodine, choline, omega3_epa_dha, melatonin, creatine, coq10, ashwagandha, collagen). For an ingredient with no matching key, still include it with its name; the engine reports it as not-yet-covered.

2. Run the engine (never invent scores)

python {skill_dir}/scripts/score.py --json '{...the product json...}'

Add --pregnant if the user is pregnant or nursing. The script prints JSON with formula, quality, safety, overall, flags, unknown, coverage, rationale, and disclaimer. To compare products, run it once per product.

3. Present the result

  • Lead with the overall and the three sub-scores.
  • Explain why using the rationale and flags, in the user's language, concisely.
  • If unknown is non-empty, say plainly that those ingredients aren't covered and the score only reflects coverage of the formula.
  • End with the disclaimer.

Rules

  • Never fabricate or hand-calculate a score — always run the script.
  • Brand-neutral: score by the data, regardless of brand.
  • Compliance: structure-function language only ("supports", "helps maintain"); never disease claims ("treats", "cures", "prevents [disease]"). Don't advise starting/stopping medication; refer symptoms to a professional.
  • Honesty: the score is a transparent methodology, not objective truth; reference values are an international baseline that varies by country.

For the full scoring logic, the honesty framing, and detailed compliance guidance, read references/methodology.md. The tunable weights and the ingredient reference table are in references/ingredients.json.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.