PromptBase
PromptBase SelectGet Select
Close icon
General
Home
Marketplace
Create
Hire
Login
Chat
Sell
Explore

Prompt Details

Model
(claude-5-sonnet)
Token size
1,618
Example input
[ARCHITECTURE_OR_SYSTEM_DESCRIPTION]: We run a single primary PostgreSQL database with no automatic failover — if it goes down, promoting a replica requires a manual process that has historically taken 45 minutes or more when we've had to do it. Separately, our application layer sits behind an auto-scaling load balancer, and that combination has successfully handled three unplanned 5x traffic spikes over the past year — each time it scaled out within about 90 seconds and scaled back down afterward with no manual intervention needed. We also have a third-party payment API integration, but we honestly don't know how it behaves if that provider has an outage on their end — we've never had it happen and haven't tested for it. [STRESS_SCENARIO]: not provided
Example output
FRAGILITY HEAT-MAP | Component | Class | Confidence | |---|---|---| | Primary database (no auto-failover) | FRAGILE | VERIFIED | | Application layer + auto-scaling load balancer | ANTIFRAGILE | VERIFIED | | Third-party payment API integration | UNTESTED | UNTESTED | Evidence basis: the database is classified FRAGILE because stress (a primary failure) directly and disproportionately harms availability — a 45-plus-minute manual recovery is a severe, disproportionate response to a single failure event, which is the definition of fragile behavior under stress, not merely "has a single point of failure" as a separate observation. The application layer is classified ANTIFRAGILE, not just Robust, because the input describes it doing more than merely surviving stress — it demonstrably adapted (scaled out) in response to three separate real stress events and returned to baseline afterward each time, which is adaptive improvement under disorder, the specific definition of antifragile used here. The payment API is UNTESTED because the input states directly that its behavior under the relevant stress (provider outage) is unknown — this is not classified as Robust by default; absence of a known failure is not evidence of resilience. VIA-NEGATIVA RECOMMENDATIONS (removal-first, ranked by single-point-of- failure risk) 1. HIGHEST PRIORITY: The primary database's lack of automatic failover is the single most significant point of failure identified. The removal-first question is not "should we add a hot standby" (an addition) but "can this single point of failure be structurally removed" — e.g., by removing the assumption that manual promotion is acceptable at all, restructuring toward an architecture where failover is not a discrete fragile event requiring 45 minutes of human process, but a property the system has by default. This is a design-level removal of a fragile dependency, not a bolt-on fix. 2. SECOND PRIORITY: The payment API integration's untested status is itself a hidden single point of failure risk — not because it's known to be fragile, but because its behavior under stress is simply unknown, which is its own category of risk. The removal-first framing here: is there unnecessary tight coupling to this specific provider that could be removed (e.g., a hard dependency on a synchronous call at checkout time that could instead be decoupled), independent of whether a backup provider is ever added. SECONDARY ADDITION NOTES (only after the above removal analysis) Only after addressing the database's structural single point of failure directly, a secondary addition worth considering is automated failover tooling — but this is explicitly secondary to the removal-first question above, not the first recommendation. For the payment integration, adding a fallback provider is a reasonable secondary step, but only after confirming through actual testing (see below) that the current integration's failure behavior is understood — adding a fallback to an unknown failure mode risks building redundancy around the wrong problem. UNTESTED COMPONENTS FLAG The payment API integration is UNTESTED. Recommendation: before deciding on any structural change, run a controlled test simulating the provider's outage (e.g., a sandboxed failure injection or a scheduled test during low-traffic hours) to actually observe behavior, rather than guessing whether it would degrade gracefully or fail hard. This observation should happen before further architectural decisions are made about this component specifically. ```
Claude

System Fragility Classifier 0

Add to Cart
Instant accessInstant access
Usage rightsCommercial use
Money-back guaranteeMoney‑back
By purchasing this prompt, you agree to our terms of service
CLAUDE-5-SONNET
Tested icon
Guide icon
4 examples icon
Free credits icon
This prompt classifies each component of your system against a fragile/robust/antifragile spectrum based on how it actually responds to stress, then applies removal-first analysis: what single point of failure or brittle dependency should come OUT, before anything gets added. It works from your description, not live telemetry, and says so plainly.
...more
Added over 1 month ago
Report
Browse Marketplace