- Home
- Skills
- Testing & Debugging
- What-If: Universal Dynamic Stress-Testing Engine
Works with the AI Agents you already use
What-If: Universal Dynamic Stress-Testing Engine
Autonomously uncover hidden failure modes, race conditions, and business logic exploits across any code, database schema, or architecture using the 5-Layer Reality Stack and 3-Tier Remediation Blueprint.
$9.99
What-If: Universal Dynamic Stress-Testing Engine
Example session with this skill installed
/what-if --deep
We are launching flash-sale VIP table reservations and coupon discounts for a high-traffic event platform. Here is our reservation SQL and checkout logic:
-- Reserve VIP table and deduct stock
UPDATE venue_tables
SET reserved_count = reserved_count + 1
WHERE id = $1 AND reserved_count < max_capacity
RETURNING id, price;
async function checkoutTable(userId: string, tableId: string, coupon?: { code: string; discountAmount: number; minSpend: number }) {
const table = await db.query(RESERVE_SQL, [tableId]);
if (!table) throw new Error("Sold out");
let finalPrice = table.price;
if (coupon && finalPrice >= coupon.minSpend) {
finalPrice = finalPrice - coupon.discountAmount;
}
await paymentGateway.charge(userId, finalPrice);
return { status: "confirmed", tableId, finalPrice };
}
- Read your context and instructions
- Compiled the what-if
Summary
Decompiled 5 hidden optimistic assumptions across your SQL reservation query and TypeScript checkout flow, uncovering
3 critical High-Blast-Radius vulnerabilities (inventory lockouts on payment failure, negative charge exploits, and single-use coupon race conditions) along with drop-in atomic fixes.
🪙 Detected Failure Currencies: Money | Data Consistency | Availability
🧬 Auto-Spawned Dimensions: Inventory Lockout, Race Condition, Promo Arbitrage, Partial Commit
🎯 1. Assumption Decompilation (Hidden Optimism)
Coupled State Mutation: Assumes paymentGateway.charge() will always succeed after venue_tables.reserved_count is incremented outside a single transaction boundary.
2.
Unbounded Coupon Arithmetic: Assumes coupon.discountAmount is strictly positive and never exceeds finalPrice.
3.
Single-Tab Execution: Assumes a user cannot trigger two concurrent checkoutTable calls with the same single-use coupon in parallel tabs.
4.
Idempotent Retries: Assumes network timeouts from paymentGateway.charge() won't cause the client to retry and reserve a second table.
⚡ 2. Top 3 Silent Killers (Prioritized by Blast Radius)
[W-01] Ghost Sold-Out Lockout (Partial Commit): UPDATE venue_tables commits immediately, then paymentGateway.charge() fails or times out. The VIP table remains permanently locked (reserved_count + 1) with zero revenue collected. →
Blast Radius: HIGH
2.
[W-02] Dual-Tab Coupon Double-Spend: User opens 2 browser tabs and submits checkout at the same millisecond using a single-use coupon. Both threads pass validation before redemption is recorded. →
Blast Radius: HIGH
3.
[W-03] Negative Settlement / Arbitrage: If coupon.discountAmount ($150) > table.price ($100), finalPrice becomes -$50, either crashing the gateway or issuing an unintended credit. →
Blast Radius: HIGH
📊 3. Full Edge-Case Matrix
| ID | Layer | Scenario | Likelihood × Impact | Blast Radius | Default Failure |
|---|---|---|---|---|---|
| W-01 | L2: Concurrency | Payment fails after UPDATE venue_tables executes | High × High | HIGH | Ghost inventory depletion; VIP tables locked without payment |
| W-02 | L2: Concurrency | Concurrent requests redeem the same coupon code | Med × High | HIGH | Single-use promo applied multiple times across orders |
| W-03 | L1: Invariant | discountAmount > table.price or negative value | Med × High | HIGH | Negative finalPrice passed to payment gateway |
| W-04 | L3: Human UX | User panic-clicks "Pay Now" 4 times during 3s gateway lag | High × Med | MED | 4 tables reserved and 4 duplicate charges initiated |
| W-05 | L5: Chaos | Gateway charge succeeds + return webhook drops + user retries | Med × High | HIGH | Customer double-charged while reserving 2 separate VIP tables |
🛡️ 4. The 3-Tier Remediation Blueprint
Tier 1: Invariants & Hard Guards (Database & Code Drop-In Fix)
Use a two-phase
TTL Hold Pattern (PENDING → CONFIRMED) with explicit row locking (FOR UPDATE) and clamped arithmetic:
async function checkoutTableSafe(
userId: string,
tableId: string,
idempotencyKey: string,
couponCode?: string
) {
// 1. Atomic DB transaction creates a 10-minute expiring hold + locks coupon
const hold = await db.transaction(async (tx) => {
const [table] = await tx.query(
`SELECT id, price, reserved_count, max_capacity
FROM venue_tables WHERE id = $1 FOR UPDATE`,
[tableId]
);
if (!table || table.reserved_count >= table.max_capacity) {
throw new Error("Sold out");
}
let discount = 0;
if (couponCode) {
const [coupon] = await tx.query(
`UPDATE coupons SET used_count = used_count + 1
WHERE code = $1 AND is_active = true AND used_count < max_uses
AND $2 >= min_spend
RETURNING discount_amount`,
[couponCode, table.price]
);
if (!coupon) throw new Error("Invalid or exhausted coupon");
discount = Math.max(0, Number(coupon.discount_amount) || 0);
}
const finalPrice = Math.round(Math.max(0, table.price - discount));
await tx.query(
`UPDATE venue_tables SET reserved_count = reserved_count + 1 WHERE id = $1`,
[tableId]
);
const [reservation] = await tx.query(
`INSERT INTO reservations (user_id, table_id, amount, status, idempotency_key, expires_at)
VALUES ($1, $2, $3, 'PENDING', $4, NOW() + INTERVAL '10 minutes')
ON CONFLICT (idempotency_key) DO NOTHING
RETURNING id, amount`,
[userId, tableId, finalPrice, idempotencyKey]
);
return reservation;
});
// 2. Charge outside DB lock using deterministic idempotency key
try {
await paymentGateway.charge(userId, hold.amount, { idempotencyKey });
await db.query(`UPDATE reservations SET status = 'CONFIRMED' WHERE id = $1`, [hold.id]);
return { status: "confirmed", reservationId: hold.id, finalPrice: hold.amount };
} catch (err) {
// Immediate rollback of held seat on deterministic payment decline
await db.query(`SELECT release_expired_or_failed_hold($1)`, [hold.id]);
throw err;
}
}
Tier 2: Resilient Interfaces (Ergonomic UX)
- Generate a client-side
UUIDv4
Idempotency Key when the checkout modal mounts so panic-clicking "Pay Now" never creates duplicate reservations.
Tier 3: Governance & Operational SOP
- Run a lightweight cron reaper every 60 seconds to release
PENDINGreservations whereexpires_at < NOW()and restorereserved_count.
Next steps
- Replace the raw
UPDATEquery with the transactionalFOR UPDATEhold pattern above. - Pass
idempotencyKeyfrom the frontend checkout session to prevent double-charging on network retries.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop Relying on Static Checklists. Break Your System Before Production Does.
Most software architectures, database schemas, and pricing models don't fail because engineers forgot a basic checklist—they fail because real-world failures are combinatorial, emergent, and multi-layered. A system that passes unit tests can still collapse when Webhook B arrives before Webhook A during a 2-second network lag while a user panic-clicks the submit button.
What-If is an open-ended, domain-agnostic failure-mode and scenario stress-testing engine adapted from high-reliability systems engineering, Fault-Tree Analysis (FTA), game theory, and pre-mortem deliberation.
Why Engineers, Architects & Strategists Use What-If
- Zero-Scenario Input Mode: You don't need to brainstorm "what could go wrong." Paste raw code, a SQL migration, an API spec, a pricing rule, or a PRD—the engine decompiles your hidden optimism and discovers non-obvious edge cases autonomously.
- The "Currency of Failure" Heuristic: Automatically detects what is irreversibly at stake (
Money,Data Consistency,Identity/Trust,Health/Safety,Availability/SLA,Attention,Compliance) and spawns domain-specific attack vectors on the fly. - The 5-Layer Reality Stack: Systematically stress-tests your artifact from micro to macro:
- Layer 1 — Computational & Invariants: Null coercion, float drift, negative boundaries, non-idempotent state mutations.
- Layer 2 — Architectural & Concurrency: Race conditions, out-of-order webhooks, split-brain states, thundering herds.
- Layer 3 — Human, Cognitive & Social: Panic resubmits, non-reading users, bystander dilemmas, clipboard corruption.
- Layer 4 — Business, Game Theory & Legal: Promo arbitrage, Goodhart's Law exploitation, referral loops, GDPR vs. audit immutability clashes.
- Layer 5 — Combinatorial Chaos (Swiss Cheese Model): Pairwise compound failures where 2–3 individually benign conditions collide simultaneously.
- Strict Actionability Gate (Zero Doom-Mongering): Every discovered flaw is paired with a concrete 3-Tier Remediation Blueprint (Hard Code/DB Guards, Resilient UX Interfaces, and Operational Telemetry SOPs).
Three Execution Modes
/what-if --quick— Fast Layer 1–2 scan for PR reviews and single functions (~500 tokens)./what-if --deep(Default) — Full 5-Layer architectural & concurrency audit for schemas, APIs, and PRDs./what-if --adversarial— Red-Team & Game Theory focus to expose loopholes in pricing, referral engines, and auth flows.
What-If performs static, architectural, and game-theoretic stress-testing on source code, SQL schemas, and system specs; it does not run live load-testing traffic or active network exploits against production servers. For large monorepos, best results are achieved by scoping audits to specific modules, transaction flows, or PRDs using frontier reasoning models (such as Claude Sonnet/Opus, GPT-4o, or Gemini Pro).
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 9 days ago
Requires