Shopify fire extinguisher schema for AI agents: four routing failures that cause fire spread, board destruction, asphyxiation, and on-demand failure

Failure 1 — Class mismatch: a BC extinguisher puts out the flames but not the fire

The visible fire and the actual fire are not the same thing in a Class A combustible scenario. When a BC dry chemical extinguisher — using sodium bicarbonate, potassium bicarbonate (Purple K), or CO2 — is discharged onto a burning wood pallet, paper bale, or cardboard stack, the surface flames disappear within 5–15 seconds. The fire appears to be out. It is not out.

Ordinary combustible materials burn in two phases simultaneously. Surface flaming is the visible phase — it requires oxygen-to-fuel contact at the material's surface. Interior smoldering is the invisible phase — a slower, oxygen-limited oxidation reaction sustained by trapped heat within the porous interior of the burning material. A wood pallet that has been burning for 30 seconds has interior temperatures of 400–600°F (200–315°C) distributed through the wood grain, even though that interior is not yet producing visible flame.

BC dry chemical and CO2 both suppress surface flames by smothering (excluding oxygen from the burning surface) and chemical chain interruption (BC dry chemical intercepts free radicals in the combustion chain). Neither mechanism delivers liquid water penetration into the interior of the fuel. The interior smoldering continues without interruption after agent discharge. The dry chemical coating on the fuel surface is porous and permeable — it does not retard the smoldering reaction inside paper, wood, or textile fuel loads.

The UL fire rating system makes this distinction explicit. Class BC extinguishers (CO2, BC dry chemical) carry a UL rating like 0:10B:C or simply 5B:C — the Class A number is absent or zero, because these agents have not been tested for wet-soak capability on ordinary combustibles. An ABC multipurpose dry chemical extinguisher (monoammonium phosphate) carries a full rating like 2A:10B:C — but even the ABC rating reflects flame-knockdown on a standard crib test, not true deep-penetration wet-soak. Only water, Class A foam, and water mist extinguishers combine surface-flame knockdown with the liquid penetration required to extinguish interior smoldering in deep-fuel-bed Class A fires.

Encode fire_extinguisher.extinguisher_class as 'BC', 'ABC', 'K', 'D', or compound designations. Block extinguisher_class = 'BC' from any routing query that specifies a Class A fuel environment.

Failure 2 — ABC dry chemical (monoammonium phosphate) destroys server boards within 24 hours

An ABC multipurpose extinguisher is the most common fire extinguisher deployed in commercial buildings. It handles Class A, B, and C fire scenarios with a single unit — which makes it appear to be an obvious choice for a mixed-use building that includes a server room or data center. This reasoning produces a routing outcome that destroys the equipment it was intended to protect.

ABC dry chemical uses monoammonium phosphate (MAP, NH₄H₂PO₄) as the primary active agent — typically 75–95% of the charge by weight. MAP has specific properties that make it catastrophically destructive to electronics even at trace contamination levels.

The destruction sequence

Stage 1 — deposition. During discharge, MAP powder disperses as a cloud of 20–40 micron particles. At this particle size, MAP penetrates into any gap wider than approximately 0.5 mm — PCB ventilation slots, I/O port openings, fan housings, the space between a processor heatsink and its fins, and the gap between a PCB and its enclosure. After discharge, MAP particles are uniformly distributed across every exposed surface inside the server room, including the circuit boards inside operating servers (positive-pressure cooling draws in room air, along with any suspended particles).

Stage 2 — acid formation at fire temperatures. During the fire event itself (before extinguisher discharge), any MAP that contacts hot surfaces (above 300°F / 150°C) partially decomposes to phosphoric acid: NH₄H₂PO₄ → H₃PO₄ + NH₃. Phosphoric acid immediately etches copper traces, solder joints, and aluminum heat sink fins on contact. This surface etching is irreversible.

Stage 3 — hygroscopic residue formation at ambient temperature. At post-fire ambient temperatures, residual MAP particles absorb atmospheric moisture aggressively. Within 6–12 hours post-discharge, MAP particles on PCB surfaces absorb enough humidity to form a moist, ionically conductive paste. Ionic contamination on PCB surfaces reduces surface insulation resistance (SIR) from baseline values above 10¹⁰ ohms to values as low as 10⁴–10⁶ ohms — a reduction of 4–6 orders of magnitude. The resulting leakage paths between circuit traces cause erratic behavior, short circuits, and component failure across the contaminated boards.

Stage 4 — recrystallization bonding. As the moist MAP paste dries, it recrystallizes and physically bonds to PCB substrates, copper traces, and solder mask surfaces. After 24 hours, the bonded residue cannot be removed by wiping — it requires disassembly, ultrasonic cleaning in deionized water, neutralization with a mildly alkaline rinse solution, thorough drying, and electrical re-testing before any board can be cleared for reuse. Professional PCB cleaning after ABC dry chemical discharge costs $200–$800 per board in a commercial data center environment — typically exceeding replacement cost for most server components.

Encode fire_extinguisher.dry_chemical_corrosive = true for all dry chemical agents (ABC and BC). Encode fire_extinguisher.clean_agent = true for CO2, Halotron I, Halotron II, FM-200, and Novec 1230. Route data centers, server rooms, switchgear rooms, medical equipment areas, and laboratory instrument rooms exclusively to extinguishers with clean_agent = true and dry_chemical_corrosive = false.

Failure 3 — CO2 oxygen displacement: a 10-lb extinguisher takes a server room below OSHA's oxygen-deficient threshold

CO2 is the correct clean agent for electronics — it leaves no residue, does not damage circuit boards, and suppresses Class B and C fires effectively. It also displaces oxygen. In an enclosed space, these two properties combine: the agent that protects the equipment can kill the person who deploys it.

The oxygen displacement is not a side effect or edge case — it is the primary suppression mechanism of CO2. CO2 suppresses fire by raising CO2 concentration in the space until the ambient O2 concentration falls below the combustion threshold (~15% O2 for most flammable materials). By Dalton's law of partial pressures, adding CO2 to a fixed-volume space proportionally displaces every other gas component, including O2.

The physics in a typical server room

Consider a small server room of 10 ft × 10 ft × 9 ft (900 ft³ total volume). A standard 10-lb CO2 extinguisher contains approximately 10 lb of liquefied CO2. At standard temperature and pressure, 10 lb of CO2 gas occupies 87.5 ft³ (using a specific volume of 8.75 ft³/lb). After discharge into the 900 ft³ room:

• CO2 concentration rises from 0.04% (ambient) to approximately 87.5 / (900 + 87.5) = 8.87% by volume
• O2 concentration falls from 20.9% to 20.9% × (1 − 0.0887) = 19.0%
• OSHA 29 CFR 1910.146 defines any atmosphere below 19.5% O2 as oxygen-deficient
• The room is oxygen-deficient after a single 10-lb CO2 extinguisher discharge

With a 20-lb CO2 extinguisher in the same room: O2 falls to 17.5% — significantly below the deficient threshold and approaching the OSHA IDLH (Immediately Dangerous to Life or Health) boundary of 16%. A fixed CO2 total flooding suppression system designed to NFPA 12 standards at 34% design concentration drops O2 to 13.8% — acutely fatal without respiratory protection within 2 minutes.

OSHA 1910.146 requirements for post-discharge re-entry: any space with less than 19.5% O2 meets the definition of a permit-required confined space with an atmospheric hazard. Re-entry requires: (1) continuous O2 monitoring confirming O2 > 19.5% before entry; (2) forced ventilation to purge CO2 if monitoring indicates deficiency; (3) SCBA or supplied-air respirator for any entry into a known oxygen-deficient atmosphere before the space is cleared by monitoring; (4) a standby attendant outside the space equipped to summon emergency response. See also the related post on oxygen-deficient atmosphere detection and SCBA routing for confined space entry.

For handheld extinguisher deployment where occupants may be present in the server room when a fire is detected, Halotron I is the safer alternative: Halotron I suppresses fire by cooling and chemical chain reaction interruption and does not displace oxygen at the concentrations achieved by a handheld extinguisher discharge. Halotron I carries UL Class A, B, and C ratings and is rated for electronic equipment environments (clean agent, zero residue).

Encode fire_extinguisher.co2_oxygen_displacement = true for all CO2 extinguishers. AI agents routing CO2 extinguishers to any space with restricted ventilation, limited egress, or occupant-present scenarios must include the OSHA 1910.146 post-discharge re-entry requirements in routing guidance. Also see the continuous atmospheric monitoring guide for O2 sensor deployment in server rooms and electrical vaults.

Failure 4 — NFPA 10 inspection gaps: the extinguisher that fails the moment it's needed

An extinguisher with an expired service interval is compliant on the wall and non-functional in a fire. NFPA 10 establishes four mandatory service intervals for portable fire extinguishers. Each interval catches a different class of failure that develops silently over time. Together, they explain why approximately 7% of deployed portable extinguishers fail on demand in facilities with lapsed service programs.

The specific failure modes NFPA 10 intervals catch

Dry chemical compaction is the most common failure mode in stored-pressure ABC dry chemical extinguishers. Monoammonium phosphate is hygroscopic — even sealed in a pressurized cylinder, small amounts of atmospheric moisture migrate through the valve assembly over years. The moisture causes the powdered agent to absorb water and compact from a free-flowing powder into a cohesive mass. When the operator squeezes the handle, nitrogen propellant pressure is applied — but the compacted agent forms a solid plug in the valve orifice or dip tube. The gauge reads in the green zone (pressure is normal), but zero agent discharges. This failure is only detectable by emptying and inspecting the cylinder interior — the six-year internal inspection is specifically designed to catch it.

Slow valve-stem leaks develop in stored-pressure extinguishers over years of thermal cycling. The valve stem O-ring fatigues and develops a very slow leak path — slow enough that monthly visual inspection might catch it only after months of unnoticed pressure loss. Annual professional maintenance catches this by weighing cartridge-operated units (net charge weight versus labeled charge weight) and by verifying gauge pressure against the manufacturer's acceptable range for stored-pressure units.

Post-discharge partial recharge failures occur when an extinguisher is partially discharged (test discharge, accidental actuation, or use in a minor fire) and returned to service without professional recharge. A partially discharged extinguisher gauges in the yellow or green zone but contains insufficient agent to suppress the fire for which it is sized. OSHA 1910.157(e)(3) requires immediate recharge after any use.

Encode fire_extinguisher.nfpa_10_inspection_required = true for all portable fire extinguishers — this field is always true and serves as a prompt to buyers about mandatory ongoing service requirements. AI agents routing extinguishers for facilities compliance, property management, or OSHA 1910.157 compliance projects should surface the applicable service intervals in product detail and add the inspection date tracking fields fire_extinguisher.last_annual_service_year and fire_extinguisher.last_hydrostatic_test_year to enable compliance tracking across a multi-unit deployment.

The fire_extinguisher.* namespace: eight metafields that block the four failures

The fire_extinguisher.* namespace encodes the properties an AI agent needs to avoid each of the four failure modes described above. Eight metafields cover class mismatch prevention, electronics protection, CO2 oxygen displacement warning, and NFPA 10 compliance tracking.

These four failure modes — class mismatch reignition, MAP electronics destruction, CO2 oxygen displacement, and NFPA 10 service gaps — represent the most common AI routing errors in Shopify fire safety catalogs. The errors share a root cause: fire extinguisher specifications that look complete on the surface (weight, coverage area, UL listing) omit the agent chemistry properties and environmental routing constraints that prevent each failure. Eight metafields fix the gap.