plugins/stripe-billing-master/skills/stripe-credit-audit-trail/SKILL.md
credit_transactions audit trail invariants and daily reconciliation patterns. PROACTIVELY activate for: (1) Designing a credit / balance / entitlement ledger, (2) credit_transactions row shape (delta, balanceAfter, reason, referenceType, referenceId, idempotencyKey), (3) Canonical idempotency key formats (refund:/debit:/stripe_checkout:/stripe_invoice:/stripe_refund:/stripe_signup_bonus:/dispute_hold:/dispute_restore:), (4) Daily reconciliation cron design (snapshot comparison, not full-sum), (5) Alert-only policy (never auto-correct drift), (6) Canonical refund helper pattern (single source of truth across pg-raw and Drizzle call sites), (7) Past_due block + credit balance shared preflight, (8) creditsDeducted boolean return pattern for pre- vs post-deduction error differentiation, (9) Money-safe email description based on resolvedVia flag, (10) Skip-users-without-baseline-snapshot reconciliation logic. Provides: table schema, canonical helper skeleton, reconciliation SQL, email gating pattern.
npx skillsauth add JosiahSiegel/claude-plugin-marketplace stripe-credit-audit-trailInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
| Field | Purpose |
|--|--|
| userId | Who the transaction belongs to |
| delta (signed int) | +granted / -revoked |
| balanceAfter (int snapshot) | Reconciliation baseline |
| reason (enum) | Categorical — one of a fixed set |
| referenceType + referenceId | Pointer to the originating entity |
| idempotencyKey (UNIQUE partial index) | Dedup + G1 checkpoint |
| metadata (jsonb) | Free-form context (previousStatus, resolvedVia, etc.) |
Pick one domain prefix that matches your billable entity (order, job, item, etc.) and use it consistently. The table below uses order as the neutral example:
| Idempotency key format | When |
|--|--|
| debit:{orderId} | order creation debit |
| refund:{orderId} | order refund |
| debit:batch:{batchId} | batch debit |
| refund_unused:batch:{batchId} | batch partial refund |
| stripe_checkout:{sessionId} | credit pack grant |
| stripe_invoice:{invoiceId} | subscription renewal |
| stripe_refund:{eventId} | Stripe refund event |
| stripe_signup_bonus:{userId} | signup bonus |
| dispute_hold:{disputeId} | dispute-initiated past_due hold |
| dispute_restore:{disputeId} | dispute closure status restore |
Use whenever any path mutates a balance / credit / entitlement column; whenever a reconciliation or audit cron is being designed; whenever a refund helper is being extended.
Related skills:
shouldRestoreStatus: stripe-billing-master:stripe-refund-dispute-lifecyclestripe-billing-master:stripe-webhook-idempotencyresolvedVia flag that gates money-safe email rendering: stripe-billing-master:stripe-list-pagination-previous-attributesIn the example below,
orders.amount_usedis a per-order ledger of credits spent on that order. Zeroing it on refund returns the spent amount (viaRETURNING), which is then re-credited tousers.credit_balancein the same transaction. Renameorders/amount_usedto match your billable-entity schema.
One helper. Two call patterns (pg-raw for Cloudflare Workers / Node workers, Drizzle for Next.js route handlers). Byte-identical SQL kept as exported constants. In the example below, orders stands in for whatever your billable-entity table is named; rename consistently across the helper, the key format, and all call sites.
// example path: packages/shared/src/refund-credits.ts
export const REFUND_ORDER_SQL = {
zeroOrder: `UPDATE orders SET amount_used = 0 WHERE id = $1 AND amount_used > 0 RETURNING amount_used`,
creditUser: `UPDATE users SET credit_balance = credit_balance + $1 WHERE id = $2 RETURNING credit_balance`,
insertTransaction: `INSERT INTO credit_transactions (user_id, delta, balance_after, reason, reference_type, reference_id, idempotency_key)
VALUES ($1, $2, $3, 'order_refunded', 'order', $4, $5)
ON CONFLICT (idempotency_key) WHERE idempotency_key IS NOT NULL DO NOTHING`,
};
type RefundArgs = { userId: string; orderId: string };
type RefundResult = { refunded: number; alreadyRefunded: boolean };
// pg-raw variant (Workers, Node workers, any pg.Client / pg.Pool compatible driver).
export async function refundOrderCreditsPg(
pg: { query: (sql: string, params: unknown[]) => Promise<{ rows: any[] }> },
{ userId, orderId }: RefundArgs,
): Promise<RefundResult> {
await pg.query("BEGIN", []);
try {
const zeroed = await pg.query(REFUND_ORDER_SQL.zeroOrder, [orderId]);
const returnedAmount: number = zeroed.rows[0]?.amount_used ?? 0;
if (returnedAmount <= 0) {
await pg.query("COMMIT", []);
return { refunded: 0, alreadyRefunded: true };
}
const credited = await pg.query(REFUND_ORDER_SQL.creditUser, [returnedAmount, userId]);
const creditedBalance: number = credited.rows[0].credit_balance;
await pg.query(REFUND_ORDER_SQL.insertTransaction, [
userId,
returnedAmount, // delta (+returnedAmount)
creditedBalance, // balance_after snapshot
orderId, // reference_id
`refund:${orderId}`, // idempotency_key
]);
await pg.query("COMMIT", []);
return { refunded: returnedAmount, alreadyRefunded: false };
} catch (err) {
await pg.query("ROLLBACK", []);
throw err;
}
}
// Drizzle variant (Next.js route handlers / any Drizzle tx). Caller passes tx + sql tag.
export async function refundOrderCreditsDrizzle(
tx: { execute: (q: unknown) => Promise<{ rows: any[] }> },
sql: { raw: (s: string, ...p: unknown[]) => unknown },
{ userId, orderId }: RefundArgs,
): Promise<RefundResult> {
const zeroed = await tx.execute(sql.raw(REFUND_ORDER_SQL.zeroOrder, orderId));
const returnedAmount: number = zeroed.rows[0]?.amount_used ?? 0;
if (returnedAmount <= 0) {
return { refunded: 0, alreadyRefunded: true };
}
const credited = await tx.execute(sql.raw(REFUND_ORDER_SQL.creditUser, returnedAmount, userId));
const creditedBalance: number = credited.rows[0].credit_balance;
await tx.execute(
sql.raw(
REFUND_ORDER_SQL.insertTransaction,
userId,
returnedAmount, // delta (+returnedAmount)
creditedBalance, // balance_after snapshot
orderId, // reference_id
`refund:${orderId}`, // idempotency_key
),
);
return { refunded: returnedAmount, alreadyRefunded: false };
}
Call sites: every service, queue consumer, and worker entry point that refunds on behalf of a user MUST delegate to this helper. Never add a new refund call site with inline SQL — that's how audit rows go missing on one branch and the reconciliation cron stops being authoritative.
-- Snapshot comparison, no full sum
WITH latest AS (
SELECT DISTINCT ON (user_id) user_id, balance_after, created_at
FROM credit_transactions
ORDER BY user_id, created_at DESC
)
SELECT u.id, u.credit_balance AS actual, l.balance_after AS expected, u.credit_balance - l.balance_after AS drift
FROM users u
JOIN latest l ON l.user_id = u.id
WHERE u.credit_balance <> l.balance_after;
Alert-only policy — never auto-correct. Users with zero credit_transactions rows since the table landed have no baseline; skip them.
creditsDeducted boolean patternSplit credit-deduction and queue-submission into separate try/catch blocks. Return shape always carries creditsDeducted: boolean so callers can distinguish:
creditsDeducted=false, caller SKIPS refundcreditsDeducted=true, caller refunds via canonical helpercreditsDeducted=true, caller refundsNever gate refunds on the order row's amount_used column — it was set BEFORE the UPDATE succeeded, so reading it post-error can mint free credits for deductions that never actually committed. Rely on the explicit creditsDeducted flag returned from the deduction function instead.
development
This skill should be used when the user asks to train, debug, scale, or improve ML models. PROACTIVELY activate for: (1) PyTorch, TensorFlow/Keras, JAX, Flax, Hugging Face Trainer/Accelerate training loops, (2) distributed training, DDP/FSDP/DeepSpeed, TPU/GPU setup, (3) mixed precision AMP/bf16, gradient accumulation, checkpointing, seeding, (4) overfitting, imbalance, loss functions, regularization, LR schedules, warmup, (5) memory optimization, gradient checkpointing, offloading, quantization-aware training. Provides: reproducible training best practices across deep learning and classical ML.
development
This skill should be used when the user asks to productionize, track, version, govern, monitor, or automate ML systems. PROACTIVELY activate for: (1) MLflow, Weights & Biases, Neptune, Comet, ClearML experiment tracking, (2) model registry, model versioning, artifact lineage, reproducibility, (3) Kubeflow, SageMaker Pipelines, Vertex AI Pipelines, Azure ML pipelines, Databricks workflows, (4) CI/CD, continuous training/evaluation, A/B tests, canary/shadow deployments, (5) drift detection, model monitoring, data validation, responsible AI governance. Provides: end-to-end MLOps architecture and operational safeguards.
development
This skill should be used when the user asks to optimize, export, serve, compress, or accelerate ML inference. PROACTIVELY activate for: (1) latency, throughput, p95/p99, batching, concurrency, KV cache, memory, or cost issues, (2) quantization INT8/INT4, GPTQ, AWQ, bitsandbytes, pruning, sparsity, distillation, (3) ONNX export, ONNX Runtime, TensorRT, TorchScript, torch.compile, XLA, OpenVINO, Core ML, TFLite, (4) Triton, TorchServe, TF Serving, BentoML, Seldon, KServe configuration, (5) edge deployment, CPU/GPU/TPU/Inferentia serving. Provides: hardware-aware inference optimization and safe benchmarking.
testing
This skill should be used when the user asks to tune hyperparameters, run sweeps, optimize search spaces, or use AutoML. PROACTIVELY activate for: (1) Optuna, Ray Tune, FLAML, AutoGluon, Hyperopt, Nevergrad, KerasTuner, W&B sweeps, (2) grid search, random search, Bayesian optimization, TPE, Gaussian processes, evolutionary search, (3) ASHA, Hyperband, successive halving, multi-fidelity optimization, population-based training, (4) learning-rate finder, batch-size search, early stopping, pruning, (5) reproducible sweep design and experiment analysis. Provides: budget-aware hyperparameter search strategy.