Shelfery Trust, Safety, and Food Claims Standard¶

This is the most important domain document in the Shelfery venture. It establishes a strong preliminary standard governing how Shelfery may represent food freshness, storage, and safety information. It applies to product content, marketing copy, design, documentation examples, and any AI-generated content related to Shelfery.

This standard is marked status: proposed because it is a strong preliminary position that still requires eventual formal approval by a role with authority to approve trust and safety policy. Until formally approved, it should nonetheless be treated as binding guidance — "proposed" here means "not yet formally ratified," not "optional."

1. Shelf-life guidance must identify authoritative sources¶

Any shelf-life, freshness, or expiration guidance presented to users must be traceable to an identified, authoritative source (e.g. a recognized food-safety authority or standards body). Shelfery must not present shelf-life numbers that were generated without a citable source. No specific authoritative source has been selected yet for this venture (see docs/strategy/assumptions-and-open-questions.md, Q5) — this is an open implementation question, and content should not proceed as if a source has already been chosen.

2. Storage guidance must distinguish conditions¶

Guidance must clearly distinguish between pantry, refrigerator, and freezer storage conditions. The same food item behaves very differently depending on where and how it is stored, and guidance that does not account for this is misleading by omission.

3. Estimates must not be represented as guarantees¶

Any freshness, shelf-life, or expiration information Shelfery shows is an estimate. It must be presented as an estimate — through language, visual treatment, and interaction design — and never as a guarantee, certification, or authoritative determination that a specific item is safe or unsafe to consume.

4. Distinguish quality guidance from safety guidance¶

The product's UX and content must clearly separate "this might not taste as good anymore" (quality) from "this might not be safe to eat" (safety). Conflating the two — for example, implying that a quality decline necessarily means a safety risk, or vice versa — is prohibited.

5. High-risk food advice requires human review¶

Any content addressing higher-risk food safety topics (e.g. perishable proteins, home canning, fermentation, or similar higher-risk categories) must be reviewed by a human with appropriate judgment before being published or presented as customer-facing. This applies regardless of whether the content was drafted by a person or generated by an AI agent.

6. Medical, allergy, nutrition, and dietary advice are out of scope¶

Shelfery does not currently provide medical, allergy, nutrition, or dietary advice, and no content should imply that it does. This category is explicitly out of scope unless and until it is deliberately developed as its own reviewed capability with appropriate expertise and review processes behind it. Do not add allergen warnings, nutritional claims, or dietary suitability statements (e.g. "safe for X diet") without that dedicated process.

7. User-entered data must remain distinguishable from authoritative guidance¶

Dates, quantities, and assumptions entered by a user (e.g. "I think I bought this last Tuesday") must remain visually and structurally distinguishable from authoritative guidance sourced from an identified authority. The product must never blend the two into an undifferentiated "fact."

8. Do not imply food is safe merely because an estimate has not elapsed¶

The absence of an elapsed estimate window must never be presented or implied as proof that a food item is currently safe or properly handled. Mishandling (e.g. temperature abuse, contamination) can make food unsafe regardless of elapsed time, and the product must not create a false sense of safety based solely on a timer not having run out.

Applicability¶

This standard applies to:

Any customer-facing product copy, UI text, or notifications related to food freshness, storage, or safety.
Marketing and brand content that references food safety or freshness.
Example data, mockups, and documentation that could be mistaken for real guidance.
AI-generated content of any of the above kinds — see docs/ai/guardrails.md for the specific prohibited behaviors this implies for AI agents.

Path to formal approval¶

Before this standard can be marked approved, it should be reviewed by a role with authority over trust, safety, and legal/compliance considerations for customer-facing food-safety content, and should likely be informed by legal review given the subject matter. Until that review occurs, treat this document as binding preliminary guidance, not as a finished, ratified policy.

Report an issue about this page