AI Unified Process - stack-agnostic core methodology (requirements, entity model, use cases)
92
95%
Does it follow best practices?
Impact
92%
1.29xAverage score across 12 eval scenarios
Low
Low-risk findings worth noting
Reverse-engineer Express.js codebase to AI Unified Process artifacts
All three artifacts
100%
100%
No /health use case
100%
100%
Aggregated use cases
40%
100%
Two actors identified
100%
100%
PlantUML format
100%
100%
UC IDs in diagram
100%
100%
Spec files match diagram
100%
100%
Business-level steps
100%
100%
Alternative flows present
50%
100%
Entity model ER diagram
0%
100%
AI Unified Process data types
12%
100%
Five-column attribute tables
0%
100%
No endpoint-named use cases
100%
100%
Business rules numbered
100%
100%
Entity model validation rules and multi-column constraints
Email validation rule
0%
100%
Decimal type for price
70%
100%
Decimal precision 10,2
0%
100%
Enrollment status uses Values rule
0%
100%
Multi-column constraint line
33%
100%
5-column attribute tables
12%
100%
No empty Validation Rules
100%
100%
Allowed data types only
25%
100%
No attributes in Mermaid diagram
50%
100%
No prose descriptions
100%
100%
Uppercase entity headings
0%
100%
Entity model ER diagram and attribute table format
Output path
100%
100%
Mermaid block present
0%
100%
No attributes in diagram
0%
100%
Correct relationship syntax
0%
100%
Entity ### headings uppercase
0%
100%
One-sentence description per entity
0%
100%
5-column attribute table
20%
100%
AI Unified Process data types only
0%
100%
Valid validation rules
0%
100%
No empty validation rules
62%
100%
Multi-column constraint note
28%
100%
No prose or Relationships table
80%
100%
PlantUML use case diagram from requirements
@startuml / @enduml
100%
100%
Left to right direction
100%
100%
Actor declaration syntax
100%
100%
Rectangle system block
100%
100%
UC-{3-digit} IDs
0%
0%
Usecase two-line format
0%
0%
Arrow syntax
100%
100%
All actors connected
100%
100%
FR coverage (no NFR/constraint use cases)
100%
100%
No implementation details in names
100%
100%
Requirements catalog format and quality checks
Correct output path
100%
100%
FR user story format
0%
100%
NFRs are measurable
100%
100%
Separate tables per type
100%
100%
FR ID prefix
100%
100%
NFR ID prefix
100%
100%
Constraint ID prefix
0%
100%
No duplicate IDs
100%
100%
Status column filled
100%
100%
NFR category column
100%
100%
Singular requirements
100%
57%
Requirements error recovery from ambiguous vision
Ambiguous NFR rewritten
100%
100%
Conflict flagged in notes
100%
100%
Conflict resolution documented
100%
100%
Missing roles noted
100%
100%
User story format
0%
0%
All NFRs measurable
100%
100%
Separate requirement tables
100%
100%
Correct ID prefixes
75%
62%
No duplicate IDs
100%
100%
Status column filled
100%
100%
Assumed threshold documented
100%
100%
Notes file structure
100%
100%
Reverse-engineer a C# / Blazor .NET 10 project into AI Unified Process artifacts
PLAN.md exists
100%
100%
Blazor entry points listed
100%
100%
Feature clustering documented
100%
100%
No C# or SQL types in entity model
75%
25%
AIUP types used
100%
70%
UC IDs grouped by actor
100%
100%
Endpoints aggregated into use cases
0%
0%
Large codebase clustering strategy for reverse engineering
PLAN.md exists
100%
100%
Controller file list
100%
100%
Feature clustering documented
100%
100%
Excluded tables listed
100%
100%
No SQL types in entity model
70%
100%
AI Unified Process types used
100%
20%
spring_session absent
100%
100%
audit_log absent
100%
100%
UC IDs grouped by actor
20%
0%
Partial implementation noted
100%
0%
Endpoints aggregated into use cases
100%
100%
Entity model from migrations
100%
100%
End-to-end test case document chaining two use cases
Output path and naming
100%
100%
Overview section
62%
100%
Roles section
0%
100%
Preconditions name their data source
100%
100%
Flow table with five columns
0%
100%
Both use cases linked
30%
100%
Journey order carries state
100%
100%
Literal test data
60%
100%
Verification steps present
50%
100%
Validation section
70%
100%
No per-use-case detail re-tested
62%
100%
No implementation detail
100%
100%
Priority in Overview
100%
100%
Postconditions cleanup inventory
75%
100%
Updating an existing use case diagram when requirements change
UC-001 preserved
100%
100%
UC-002 preserved
100%
100%
UC-003 preserved
100%
100%
No ID gaps or duplicates
100%
100%
FR-004 use case added
100%
100%
FR-005 use case added
100%
100%
Auditor actor declared
100%
100%
Auditor actor connected
100%
100%
Valid PlantUML structure
100%
100%
No implementation details in names
100%
100%
All actors connected
100%
100%
Use case spec completeness and structure
File location & name
100%
100%
Overview: UC-XXX ID
100%
100%
Overview: primary actor
100%
100%
Overview: goal stated
71%
100%
Overview: valid Status
100%
100%
Numbered steps, no gaps
100%
100%
Main scenario ends with goal
85%
100%
At least two alternative flows
100%
100%
Trigger references a step number
87%
75%
Alternative flow endings
100%
100%
Success postcondition
100%
100%
Failure postcondition
100%
100%
Business rules with BR-XXX IDs
85%
100%
Business-level step language
87%
100%
Business-level use case step writing and multi-file use cases
UC-001 file name
0%
100%
UC-002 file name
0%
100%
One use case per file
100%
100%
No SMTP/email-protocol steps
0%
100%
No password-hashing steps
0%
100%
No JWT/token-issuance steps
0%
100%
No SQL/database steps
100%
100%
Alt flow trigger references step number
80%
100%
Alt flow ending
70%
100%
At least one alt flow per use case
100%
100%
Business rule IDs unique globally
100%
100%
Business rules present
100%
100%
Main scenario numbered steps
100%
100%
Table of Contents