> ## Documentation Index > Fetch the complete documentation index at: https://docs.corsa.finance/llms.txt > Use this file to discover all available pages before exploring further. # Rule Spec Drafting Skill > AI skill for turning a compliance policy into a precise, buildable Corsa rule description — scoped strictly to what the rule engine can express. ## Overview The **corsa-rule-spec-drafting** skill turns a **compliance policy** (a clause, obligation, or AML/CFT requirement) into a **1–2 paragraph rule description** that a developer — or the **corsa-rule-engine-authoring** skill — can implement directly. It knows exactly what Corsa's transaction-monitoring rule engine can and cannot express, so it never asks for features that don't exist and always flags capability gaps explicitly. ### What It Helps With * Converting a compliance policy clause into an implementation-ready rule spec * Scoping AML/CFT requirements for the Corsa rule engine * Identifying which policy requirements fall outside the rule engine's capability envelope * Producing grounded descriptions that hand off cleanly to rule authoring *** ## Quick Start After installing the skill, ask your coding tool questions like: * *"Write a rule spec for: 'flag senders who make 3+ deposits below \$10K in 24 hours'"* * *"Translate AML Policy §4.3 on structuring into a Corsa rule description"* * *"Can the rule engine do a live OFAC lookup? If not, what's the alternative?"* * *"Scope this EDD requirement for the Corsa rule engine: clients risk-rated HIGH must be re-screened"* *** ## What the Skill Knows ### The Two-Step Workflow ``` corsa-rule-spec-drafting → grounded rule description corsa-rule-engine-authoring → valid rule JSON + deployment ``` Use this skill first to produce the spec, then hand it to **corsa-rule-engine-authoring** for implementation. ### What Rules Can Express * **Simple comparisons** on transaction, client, wallet, and bank account fields — amount, type, status, risk tier, sanctions/PEP flags, country, KYC tier, and more * **Rolling time-window aggregations** — cumulative sum, count, distinct count, velocity change, standard deviation, baseline comparison * **Nested AND/OR logic** with conditions scoped to sender, receiver, or both * **Dynamic thresholds** comparing against a client's configured daily or monthly limit * **Actions**: create an alert (with category, priority, and assignment) and/or freeze the transaction ### What Rules Cannot Express (Capability Gaps) The skill explicitly flags these so specs stay buildable: | Policy requirement | Corsa equivalent / gap | | --------------------------------------- | ----------------------------------------------------------------- | | Live OFAC/PEP lookup at evaluation time | Pre-computed sanctions/PEP flags on the client entity | | Regex or fuzzy text matching | Exact, set, and range comparisons only | | ML models or custom anomaly scoring | Built-in statistical operators (`stddev`, `deviationFromAverage`) | | Multi-hop graph / fund tracing | Only direct transaction participants are visible | | Scheduled or periodic rule runs | Use Workflows for periodic checks | | Actions beyond alert + freeze | Emails, webhooks, case creation require Workflows | ### Output Format A 1–2 paragraph plain-English description containing: the compliance basis (policy citation), the specific signals and thresholds, participant scope (sender/receiver/both), boolean logic, the action, and any explicit gaps with nearest approximations. No JSON, no field names — just intent, ready for a developer to implement. *** ## Source View the full skill source with policy-to-capability mappings and worked examples.