> ## Documentation Index
> Fetch the complete documentation index at: https://developer.watson-orchestrate.ibm.com/llms.txt
> Use this file to discover all available pages before exploring further.

# wxo-analyzer skill

The `wxo-analyzer` skill reverse-engineers existing watsonx Orchestrate projects to generate comprehensive documentation sets, component relationship maps, and anti-pattern reports across 15 enterprise AI failure modes.

**Source location:** [`skills/wxo-analyzer/SKILL.md`](https://github.com/IBM/ibm-watsonx-orchestrate-adk/blob/main/skills/wxo-analyzer/SKILL.md) in the ADK repository.

## Overview

As projects scale, understanding dependencies across multiple agents, tools, workflows, and connections becomes critical for auditing, security reviews, and developer onboarding. The `wxo-analyzer` skill scans your repository without modifying code to produce multi-level documentation and architectural health checks.

**Best for:** Auditing existing solutions, project onboarding, architecture reviews, and pre-production readiness assessments.

## Documentation deliverables

The skill generates a structured three-report documentation set:

1. **Solution Overview Report** — High-level solution topology, executive architecture summary, and full component inventory.
2. **Agent Analysis Report** — Deep dive into each agent's purpose, instructions complexity, tool bindings, context handling, and collaborator relationships.
3. **Tools & Components Report** — Detailed catalog of all Python tools, OpenAPI plugins, flows, knowledge bases, and external connections with parameter schemas and permission levels.

## Anti-pattern detection

The skill analyzes your project against 15 common enterprise AI anti-patterns:

| Category | Checked anti-patterns |
| - | - |
| **Architecture & Scope** | Overloaded "God Agent", redundant collaborator hops, missing deterministic routing boundaries |
| **Tooling & Scalability** | Unbounded tool exposure, missing type definitions, unvalidated inputs, permission over-provisioning |
| **Instruction Quality** | Hidden state assumptions, conflicting priority rules, brittle keyword matching, excessive instruction length |
| **Knowledge & Context** | Unindexed retrieval lookups, oversized context injection, missing chunking strategies |

## Relationship mapping

The skill generates Mermaid architecture and sequence diagrams depicting exact interactions:

```mermaid theme={null}
graph TD
    Client[Webchat Client] --> Orchestrator[Triage Agent]
    Orchestrator -->|Collaborator| BillingAgent[Billing Specialist Agent]
    Orchestrator -->|Collaborator| TechAgent[Technical Support Agent]
    BillingAgent --> InvoiceTool[lookup_invoice tool]
    BillingAgent --> PaymentFlow[process_refund workflow]
    TechAgent --> KB[(Product Knowledge Base)]
```

## Activation

In Bob, activate the skill by pointing it to your project folder or pasting project artifacts:

```text theme={null}
Analyze my current watsonx Orchestrate project and generate an architecture report with anti-pattern analysis.
```

Or for targeted audits:

```text theme={null}
Use wxo-analyzer to map the relationships between all agents and tools in this repository.
```

## Next steps

* For deep workflow-specific latency checks, run **[Agentic Workflow Advisor](/bob_skills/agentic_workflow_advisor)** on any exported flow JSON files.
* For instruction-level reliability auditing on a specific agent, use **[agent-instructions-evaluator](/bob_skills/agent_instructions_evaluator)**.