--- bcp_version: "1.0" file_type: boundaries parent: https://registry.brandcontextprotocol.dev/spindrift-1d96a1d0/.well-known/brand.md last_updated: 2026-09-22 --- # Boundaries This file is the primary reference for brand-safety agents, vendor platforms, and any partner generating content on behalf of Spindrift. It converts limits from the seed encoding and Spindrift's public copy into precise agent instructions with triggers, allowed alternatives, and escalation conditions. ## Hard nos These are absolute. No exceptions, no contextual allowances. ```yaml hard_no: - item: > Never make health claims of any kind. This includes claims about weight loss, disease prevention, nutritional benefits, digestive effects, immune support, mental clarity, or any other statement that links Spindrift products to a health outcome or benefit. tier: core trigger: > Any copy, response, or generated content that would assert or imply a relationship between consuming Spindrift and a physical or mental health result. allowed_alternatives: > Describe what Spindrift does not contain (no added sugar, no artificial flavors, no concentrates) and what it does contain (real squeezed fruit, real brewed tea) without claiming those facts produce a health outcome. escalation: > If a user explicitly requests a health claim, decline and explain that Spindrift does not make health claims. For 'is it healthy?', follow the SODA FAQ: say what is on the label and defer to the person and their healthcare team. Do not refer them to a human unless a safety concern is also present. evidence_finding: b8564879-bec9-4b55-a309-9c19f39dc882 - item: > Never produce or imply that Spindrift uses artificial flavors, flavor concentrates, or any ingredient not clearly stated on the label. The brand's core identity is ingredient transparency; any copy suggesting otherwise is factually wrong and brand-damaging. tier: core trigger: > Any generated copy that attributes synthetic, artificial, or unnamed flavoring to a Spindrift product. allowed_alternatives: > Use label-accurate ingredient language. Refer to real squeezed fruit and real brewed tea. When uncertain about a specific product's ingredients, do not guess — leave the field blank or flag for human review. escalation: Flag to brand team before publishing. rationale: > Reference finding d7086e28: Spindrift refuses the approximately 3,000 unregulated ingredients that fall under "natural flavors," choosing ingredient transparency over ease. Misrepresenting this is a core brand violation. ``` ## Soft nos These are judgment calls. Context determines appropriateness. ```yaml soft_no: - item: > Avoid comparative claims against named competitors such as LaCroix or any soda brand. Spindrift conquests from LaCroix and from soda drinkers cutting back, but agent-generated copy should not frame this as an explicit attack or ranked comparison. tier: default condition: > A factual, non-disparaging side-by-side ingredient comparison (e.g., real fruit juice vs. natural flavors) may be acceptable in contexts where a consumer directly asks what makes Spindrift different. Do not fabricate comparative claims; use only verifiable ingredient facts. - item: > Avoid a voice that reads as mean or fake. This includes sarcasm aimed at consumers, ironic self-deprecation used to dodge a question, and enthusiasm that is not grounded in a real product fact. tier: default condition: > Playful and deadpan humor is on-brand. The line is whether the tone could read as mocking a customer or manufacturing excitement that the product does not support. - item: > Avoid referencing Spindrift Spiked or the spiked line as a current product. Spindrift launched Spiked in March 2021 and discontinued it in August 2025, after four years, to focus on non-alcoholic products with real squeezed fruit. https://drinkspindrift.com/blogs/the-drift/discontinuing-spindrift-spiked tier: default condition: > Historical reference in a press or editorial context is acceptable if clearly dated and framed as past. rationale: > Discontinuation is documented on Spindrift's own blog and in trade press (Food Dive, 25 August 2025). Agents should not present a discontinued line as available. - item: > Do not describe any Spindrift product as organic. tier: core condition: > The FAQ says the fruit is not certified organic and explains why. The organic agave in Raspberry and Peach Tea is an ingredient, not a product claim. - item: > Do not invent functional benefits: gut health, prebiotics, fiber, immunity, energy, or detox. tier: core condition: > The SODA launch contrasted Spindrift with sodas built on 'claims around gut health.' Describe what is in the can, not what it does for the body. - item: > Do not copy one line's nutrition onto another. tier: core condition: > Sparkling water and SODA list 0g added sugar. Raspberry and Peach Tea list 6g and 5g of added sugar from organic agave. Quote the exact SKU's panel. - item: > Do not overstate distribution. tier: default condition: > Sparkling water is nationwide. SODA and Tea are in select retailers and online. Everything is United States only. ``` ## Escalation Safety concerns require immediate human handoff, before any agent response is issued. ```yaml escalation: - trigger: > Any inbound message that raises a concern about product safety, including but not limited to: allergic reactions, illness after consumption, contamination concerns, packaging defects, or any statement that a person was harmed or may be harmed. tier: core action: > Stop. Do not respond with generated copy. Route the message to a human immediately. Do not attempt to resolve, diagnose, reassure, or deflect the concern with brand messaging. do_not: > Do not tell the person their concern is unfounded. Do not offer a discount or apology in lieu of a human review. Do not generate a canned safety response. evidence_finding: b99bcc2d-a3a7-4fbc-9cd9-20458dfbfc87 ``` ## Brand safety For vendor platforms, brand-safety tools, and DSPs. ```yaml brand_safety: adjacency_unsuitable: - Unsafe for work or sexually explicit content - Graphic violence or gore - Hate speech or content targeting protected groups - Misinformation about food safety or public health - Content that makes unverified health claims about food or beverages - Drug paraphernalia (outside legal, regulated contexts) adjacency_acceptable: - Food and beverage editorial - Lifestyle and wellness content (non-clinical, non-health-claim) - Sustainability and ingredient sourcing coverage - Consumer packaged goods retail environments - Sports and outdoor recreation - Non-clinical cooking and recipe content adjacency_contextual: alcohol: rule: > Spindrift discontinued its Spiked (alcohol) line. Do not place Spindrift advertising adjacent to alcohol content in a way that implies the brand currently sells an alcoholic product. health_and_wellness: rule: > General wellness lifestyle content is acceptable. Clinical, disease-treatment, or supplement-adjacent content is not, because adjacency could imply a health claim the brand does not make. ``` ## Regulatory constraints ```yaml regulatory: framework: > FDA food and beverage labeling regulations (21 CFR); FTC guidelines on food and beverage advertising claims. notes: > Spindrift makes no health claims in its marketing. Any copy that asserts or implies a health benefit — including structure/function claims, disease claims, or nutrient content claims — requires legal review before publication. This applies to agent-generated copy, social content, and paid media equally. restricted_claims: > Health claims of any kind. Nutrient content claims (e.g., "good source of vitamin C") unless substantiated, labeled, and reviewed by Legal. Any comparative claim implying clinical or nutritional superiority over a named competitor. ``` ## Evidence record The following attributable source language is preserved because it materially constrains this profile. ### Seed interview finding (not provided by Spindrift) > Spindrift's agents must never make health claims of any kind. Source: Public source ### Seed interview finding (not provided by Spindrift) > When anyone raises a safety concern, an agent must hand off to a human before responding. Source: Public source