- Home
- Skills
- Agents & Orchestration
- AI Agent Decision Logic Builder – Decision Tree & Workflow
Works with the AI Agents you already use
AI Agent Decision Logic Builder – Decision Tree & Workflow
Convert business requirements and policies into explicit, testable decision logic and flowcharts for AI agents.
$23.99
AI Agent Decision Logic Builder – Decision Tree & Workflow
Example session with this skill installed
Design a decision tree for a refund agent. Customers get full refunds if the item is damaged and reported within 30 days. Late reports need manager approval. High-value items over $500 always need approval.
- Read your context and instructions
- Compiled the ai agent decision
I have converted your refund policy into a structured decision architecture.
AI Agent Decision Logic: Refund Processing
1. Input Catalogue
| Field | Type | Required | Validation | Missing Data Handling |
|---|---|---|---|---|
item_condition | Enum | Yes | {DAMAGED, DEFECTIVE, WRONG_ITEM, UNWANTED} | Fail to escalation |
report_days | Integer | Yes | >= 0 | Assume > 30 (Escalate) |
item_value | Decimal | Yes | > 0.00 | Assume > 500 (Escalate) |
2. Business Rule Catalogue
R1 (Auto-Refund): If item_condition IS DAMAGED AND report_days <= 30 AND item_value <= 500 -> Outcome: AUTO_APPROVE.
- R2 (Late Report): If
item_conditionIS DAMAGED ANDreport_days> 30 -> Outcome: MANAGER_REVIEW. - R3 (High Value): If
item_value> 500 -> Outcome: MANAGER_REVIEW (Priority Overwrite). - R4 (Default): If no conditions match -> Outcome: DENY_AND_ESCALATE.
3. Implementation Specification (Pseudocode)
def evaluate_refund(request):
# R3: High Value Check (Override)
if request.item_value > 500:
return {"action": "ESCALATE", "reason": "HIGH_VALUE_THRESHOLD"}
# R1 & R2: Damage Assessment
if request.item_condition == "DAMAGED":
if request.report_days <= 30:
return {"action": "APPROVE", "reason": "STANDARD_DAMAGED_POLICY"}
else:
return {"action": "ESCALATE", "reason": "LATE_REPORT_APPROVAL"}
# R4: Fallback
return {"action": "DENY", "reason": "OUTSIDE_POLICY_SCOPE"}
4. Test Scenarios
- Scenario A (Happy Path): Damaged, 10 days, $45. Expected:
APPROVE. - Scenario B (Threshold): Damaged, 10 days, $501. Expected:
ESCALATE. - Scenario C (Late): Damaged, 45 days, $50. Expected:
ESCALATE. - Scenario D (Unknown Value): Damaged, 10 days, NULL. Expected:
ESCALATE.
Next steps
- Define the specific "Manager Review" queue destination for escalated cases.
- Confirm if "Defective" or "Wrong Item" should follow the "Damaged" branch.
- Integrate the pseudocode logic into your agent's routing tool.
Connects securely to your tools. The creator never sees your data.
About this skill
AI agents often fail because their decision logic is buried in vague natural-language prompts. This makes them unpredictable, hard to debug, and prone to "hallucinating" business policies when they hit a blank spot. If your workflows break when data is missing or rules overlap, you need a structured logic specification.
This skill converts messy business requirements, policies, and user stories into implementation-ready decision architectures. It forces you to define exactly how an agent should handle every branch, from valid inputs to edge-case failures. It replaces "when appropriate" with testable predicates, ensuring your agents act as reliable extensions of your business logic.
What it does
- Rule normalization translates plain-language policies into structured, testable predicates with defined operators.
- Input cataloging defines field types, validation rules, and specific behaviors for null or malformed data.
- Conflict resolution identifies overlapping or contradictory rules and applies explicit priority policies.
- Fallback design maps out escalation paths, safe default outcomes, and human-in-the-loop approval gates.
- Implementation specs generates framework-neutral pseudocode, Mermaid flowcharts, and JSON-style rule objects.
How it works
- Intake requirements by providing existing prompts, policy docs, or workflow notes to the agent.
- Review the extraction of decision boundaries, input fields, and potential rule conflicts identified by the skill.
- Refine thresholds by answering specific questions about missing data behaviors or priority overrides.
- Export specifications in your preferred format, such as a full logic architecture or a concise decision tree.
Frameworks & tools
This skill produces framework-neutral logic applicable to LangGraph, CrewAI, PydanticAI, or custom Python/TypeScript orchestration layers. It uses Mermaid.js for diagrams and standard JSON for rule schemas.
Why this beats prompting it yourself
Writing "be logical" in a system prompt results in inconsistent behavior. This skill uses a systematic intake process to catch logical gaps, unreachable branches, and silent failures that standard prompting misses. It ensures every decision point has a defined outcome and a recovery path.
Use cases
- Customer Support Triage to route tickets based on sentiment, account tier, and specific keyword triggers.
- Financial Processing for building loan or credit approval trees with strict regulatory boundaries.
- Content Moderation to define explicit "keep" or "delete" logic based on evolving community guidelines.
- DevOps Automation for determining when to auto-remediate infrastructure alerts versus escalating to on-call.
Known limitations
This skill designs logic and does not execute code or verify production correctness. It requires the user to provide specific business thresholds and cannot invent legal or medical policies.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Does not execute workflows. Requires user-defined thresholds for business-critical values. Cannot verify legal or regulatory compliance of provided rules.
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 buying a skill.
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 17 days ago
Permissions
No special permissions declared or detected