Reviving Praderas (Day 1): technical audit and AI-agent improvement plan
Some projects do not need to be rebuilt from scratch. They need a second life.
Today we officially started that second life for Praderas. The goal is not just to "touch up" a blog, but to turn it into a living, maintainable project that keeps getting more useful for readers. A key part of the process is using AI agents to iterate faster and with better consistency.
What we did today (and why it matters)
Before writing new code, we ran a full audit on three fronts:
-
Repository and internal architecture
We reviewed PicoCMS configuration, theme structure, plugins, and content layout to understand how everything fits together. -
Real published-site behavior
We compared what the code suggests with what visitors actually see in production. -
Content and metadata quality
We evaluated editorial consistency (dates, tags, templates, and organization) to prepare sustainable improvements.
This step is essential: without clear context, technical fixes stay fragile.
Main findings
The audit surfaced several important points:
- There is a canonical-domain mismatch between expected and active behavior.
- The main blog-listing template showed signs of malformed markup in final render.
- The interface mixed Spanish and English text in key navigation areas.
- Search and pagination needed refinement for clearer navigation.
- The content base was on the right track, but metadata standardization was still incomplete.
None of this is dramatic; it is typical debt in projects that evolved in stages.
What we left ready for faster next iterations
As the last change in this first PR, we created two documents under .agents so future agents (and humans) start from shared context:
repo-context.md: project map, architecture, plugins, conventions, and current risks.proposed-improvements.md: prioritized backlog for structure, navigation, and usability, with phases and metrics.
In practical terms: now we have a compass, not just enthusiasm.
What comes next: a parallel technical log
Alongside this blog rebuild, we are starting a parallel technical log around a challenge many teams know well:
how an agent can help iteratively move a legacy PHP5 app toward modern PHP, without breaking everything in the process.
This will not be theory. We will document real decisions, trade-offs, mistakes, validation, and outcomes.
Day 1 closing
Today was not about visible polish. It was about structural clarity so we can improve with method.
Tomorrow we do not start from zero anymore. We start with context, priorities, and a plan.