Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/ronaldjdev/forge/llms.txt

Use this file to discover all available pages before exploring further.

forge assay produces a qualitative, multi-perspective architectural essay on top of the quantitative data from forge inspect. Instead of a single point of view, it convenes five simulated expert personas — each with a different professional lens — and asks every one of them to weigh in on your codebase simultaneously. The result is a richer, more actionable diagnosis than any single audit score can give you: you learn not just what is wrong, but why it matters from a scalability standpoint, a technical-debt standpoint, a security standpoint, a product standpoint, and a governance standpoint — all at once.

​
Usage

# Via agent natural language: "examinar", "calidad", "opinión", "critique"
# Direct invocation:
forge assay                     # Full review (all 5 personas)
forge assay --persona=bezos     # Only Bezos perspective
forge assay --persona=fowler    # Only Fowler perspective
forge assay --persona=hacker    # Only hacker/security perspective
forge assay --persona=pm        # Only PM perspective
forge assay --persona=senior    # Only senior architect perspective
forge assay --save              # Persist results in .forge/assay/
forge assay --json              # JSON output

​
The Five Personas

Each persona targets a distinct area of architectural concern. Running the full assay surfaces blind spots that any single reviewer would miss.
PersonaFocus
Jeff BezosCoupling, scalability, team autonomy (API mandate)
Martin FowlerRefactoring, technical debt, microservices vs monoliths
HackerSecurity, performance, edge cases, vulnerabilities
Alex (PM)Time-to-market, complexity, technical ROI
Dr. Carter (Architect)Dependency cycles, layer violations, structural health
The essay is driven by the last forge inspect result. If no prior audit exists, assay still runs but with reduced context. For the richest output, always run forge inspect first.

​
Flags

--persona=<id>

Filter to a single persona. Valid IDs: bezos, fowler, hacker, pm, senior.

--save

Persist the full assay report to the .forge/assay/ directory for later review and comparison.

--json

Emit machine-readable JSON output — useful for tooling, CI pipelines, or dashboards.

​
Saving Results

When you pass --save, Forge writes the full report to .forge/assay/ using a timestamped filename (e.g. assay-2026-06-25T00-00-00.md). Results accumulate over time, so you can track how each persona’s assessment evolves across sprints or refactoring cycles. 
forge assay --save              # Saves current assay to .forge/assay/
The .forge/assay/ directory acts as a lightweight architectural journal. Before major decisions — migrating to microservices, introducing a new feature boundary, upgrading an ORM — run assay --save so you have a recorded baseline to compare against after the change.
1

Audit first

Run forge inspect to get a quantitative score (0–100), violation list, and architecture graph.
2

Run assay

Run forge assay to get qualitative analysis from all five personas, grounded in the inspect data.
3

Prioritise

Use the personas’ recommendations to decide which fixes deliver the most value. CRITICAL violations flagged by Dr. Carter and Jeff Bezos typically block team autonomy and introduce the highest risk.
4

Refactor

Address findings with reforge, temper, or smelt as appropriate.
5

Repeat

Re-run forge inspect followed by forge assay to verify improvement and capture the new baseline with --save.
Run assay before major architectural decisions — introducing a new service boundary, splitting a monolith, onboarding a new team — to get a balanced view from multiple expert lenses before committing to a direction.

Build docs developers (and LLMs) love

Get started for freeTalk to us