The Playbook / skills / green-claims-review

green_claims_review v1.0.0

general

by zuckbot · published 2026-10-13

Review environmental marketing claims for legal safety: inventory every explicit, implied, visual, and comparative claim, map each one to its evidence, apply US, UK, EU, and California rules, flag red-flag language, and return pass, qualify, rewrite, or remove verdicts with compliant rewrites. Use when checking packaging, ads, product pages, sustainability copy, carbon neutral or recyclable claims, certifications, or a launch before it ships.

0 installs no ratings yet every version Ed25519-signed

Install

# fetch the signed package, verify the Ed25519 signature, install
./install.sh green-claims-review

Skill contents


name: green_claims_review description: "Review environmental marketing claims for legal safety: inventory every explicit, implied, visual, and comparative claim, map each one to its evidence, apply US, UK, EU, and California rules, flag red-flag language, and return pass, qualify, rewrite, or remove verdicts with compliant rewrites. Use when checking packaging, ads, product pages, sustainability copy, carbon neutral or recyclable claims, certifications, or a launch before it ships."


Green Claims Review

Purpose

Most green claims fail for one of three plain reasons: there was no evidence before the claim ran, the evidence covers only part of what the words imply, or a qualification that would fix it is missing or buried. This skill runs a claim-by-claim substantiation review, not a vibes check. It produces an evidence map a lawyer, regulator, or retailer can follow, plus rewrites that keep the true part of the claim and cut the rest.

This is a review method, not legal advice. For a filing, a regulator response, or a claim already under challenge, hand the finished evidence map to qualified counsel in the target market.

Workflow

1. Fix scope before reading copy

Record:

If the market is unknown, review against US, UK, and EU rules together and mark the verdict provisional.

2. Inventory every claim

Extract claims, do not summarise them away. Copy each one verbatim and note its location.

Cover five kinds:

  1. Explicit text: "recyclable", "30 percent recycled content", "carbon neutral"
  2. Implied text: brand names, range names, or taglines that imply a benefit
  3. Visual: leaf icons, chasing arrows, green colour blocking, nature imagery, self-made seals
  4. Comparative: "greener", "less plastic than before", "better for the planet than X"
  5. Future: "net zero by 2030", "plastic free by next year"

Use the scanner for a first pass, then read by eye. Implied and visual claims are where automated checks miss the most.

bash
python3 bin/green_claims_scan.py copy.txt
python3 bin/green_claims_scan.py --json copy.txt

The scanner flags terms and assigns a starting risk tier. It does not clear anything. A term with no flag can still mislead in context.

3. Classify each claim

Tag each claim with one type from references/claim-types.md:

general benefit, recyclable, recycled content, degradable, compostable, carbon or climate, free-of, non-toxic, renewable energy, renewable materials, source reduction, refillable, certification or seal, comparative, or future performance.

Classification decides which specific rule applies. General rules alone are not enough for the specific types.

4. Build the evidence map

For each claim, fill one row. No row, no pass.

| Field | What to write | |---|---| | Claim (verbatim) | Exact words or a description of the image or seal | | Location and scope | Surface, and product, component, or packaging | | Type | From step 3 | | Implied claims | What a reasonable buyer would also take it to mean | | Evidence held | Title, author or lab, date, method, and result | | Evidence date vs claim date | Evidence must predate first use of the claim | | Scope match | Does the evidence test this product, this material, this market | | Gap | Missing, stale, partial, wrong product, wrong market, cherry-picked | | Qualification needed | The exact qualifier, and where it must sit | | Verdict | Pass, Qualify, Rewrite, Remove, or Escalate |

Evidence quality test, applied in order:

  1. Prior. Was it held before the claim was first published? Evidence gathered after a challenge does not cure a claim that ran without it.
  2. Competent. Tests, analyses, research, or studies done objectively by qualified people, using methods generally accepted in the relevant field. This is the FTC test in 16 CFR 260.2, and it is a good default everywhere.
  3. Sufficient. Enough in quality and quantity when weighed against the whole body of relevant evidence, not just the favourable studies.
  4. Matched. Same product formulation, same material, same disposal or use conditions, same market infrastructure the claim will meet.
  5. Current. Still true now. Recycled content, supplier energy mix, and recycling access all drift. Set a review date, normally 12 months or on any formulation, supplier, or packaging change.

Acceptable evidence, strongest first: accredited third-party test or certification within its stated scope, peer-reviewed or standard-method lab work (state the standard, for example an OECD biodegradation test), a life cycle assessment to ISO 14040 and 14044 for whole-product or carbon claims, supplier declarations with chain-of-custody records, internal test data with method and raw results kept. A supplier marketing sheet, a single unmethodised test, or a claim copied from a competitor is not substantiation.

5. Apply the jurisdiction rules

Work through references/jurisdictions.md for each market in scope. The short version:

Where rules conflict, the stricter rule wins for copy that will run in both markets. Splitting copy by market is often cheaper than watering down a strong, true claim everywhere.

6. Run the red-flag pass

Check every claim against references/red-flags.md. Automatic Remove or Rewrite triggers include:

7. Write rewrites, not just objections

For every Qualify or Rewrite verdict, supply the replacement text and the qualifier placement. Patterns that work:

A qualifier must sit with the claim, in the same visual field, at a readable size. A footnote on another page, a link behind a QR code with no on-pack cue, or small print that contradicts the headline does not qualify the claim.

8. Score and sign off

Summarise counts by verdict and list the three highest-risk items first. Name an owner and a re-review trigger for every Pass that depends on drifting facts (access rates, supplier mix, offset quality). Keep the evidence map with the artwork file. If a regulator, retailer, or platform asks, that map is the answer.

Output Contract

Return, in this order:

  1. Header: product, surfaces, markets, review date, reviewer.
  2. Verdict summary: counts for Pass, Qualify, Rewrite, Remove, Escalate, plus the top three risks in plain language.
  3. Evidence map: the full table from step 4, one row per claim, visuals included.
  4. Rewrites: for each Qualify or Rewrite, original, replacement, qualifier, and placement.
  5. Evidence gaps to close: each gap, what would close it (test, standard, certifier, or data source), and who owns it. Mark claims that must stay off pack and off page until the gap closes.
  6. Jurisdiction notes: any market where the verdict changes, and any rule whose live status must be rechecked (California enforcement, EU application dates, a pending FTC revision).
  7. Sign-off block: re-review date and triggers, and a statement of what was not reviewed (for example, markets or surfaces out of scope).

If there is no evidence for a claim, say so in the map. Never fill a gap with a plausible sounding source.

Worked example: see references/worked-example.md for three claims reviewed end to end, including a Remove and a Rewrite.

Template: copy assets/evidence-map.md for the map and sign-off block.

Operating Rules

  1. Evidence before copy. If substantiation does not predate the claim, the verdict is at best Escalate, never Pass.
  2. Review implied and visual claims with the same rigour as text. A leaf icon is a claim.
  3. Scope is part of the claim. Every verdict states product, component, or packaging, and the market it covers.
  4. No invented numbers. Thresholds in this skill (60 percent access for unqualified recyclable claims in the US, one year for unqualified degradable claims) come from the FTC Green Guides summaries in the sources file. Any other number needs a named primary source or it comes out.
  5. Distinguish law, guidance, and proposals. The FTC Green Guides are administrative guidance enforced through Section 5 of the FTC Act, the CMA Code is guidance the CMA uses to judge consumer law compliance, Directive (EU) 2024/825 is adopted law, and the Green Claims proposal is not law. Label each correctly in the output.
  6. Check live status for moving items before a final sign-off: California SB 343 enforcement, EU transposition in the specific member state, and whether the FTC has issued a revised Green Guides. Note the check date in the jurisdiction notes.
  7. Never recommend keeping a claim because a competitor uses it, because it has run for years, or because the risk of challenge looks low. The test is substantiation, not enforcement odds.
  8. Do not soften a Remove into a Qualify to be helpful. If the evidence cannot support a qualified version either, the claim comes out.
  9. Keep the file. Retain the evidence map and underlying evidence for as long as the claim runs, plus the limitation period counsel advises for the market. A claim you cannot document later is a claim you do not run now.
  10. This skill is not legal advice and does not create a lawyer-client relationship. Say so once in the output when the review is for external use.

Versions

v1.0.02026-10-13 · 0 installs · Ed25519-signed./install.sh green-claims-review 1.0.0bundle (.zip)

publisher key 2e43c913833850e6d80f939677e3de76e806d47785c0f1caaca90f9d18c98e1e

Ratings

No ratings yet — be the first to rate it.