mafralp@ai-usage:~$
AI usage & agent workflows
I use AI as an integrated delivery workflow: defining the objective, grounding work in product context, selecting a specialized skill or tool, reviewing the output and shipping editable artifacts or production code. The goal is not one-off generation; it is repeatable, reviewable work across engineering, product and communication.
Codex agents
Codex is used as an implementation agent for scoped development, debugging, repository analysis, test-oriented changes and technical documentation. It works with local code, command-line validation and browser/UI automation when a task needs end-to-end verification.
Custom skills
- Directed vBuz email — recipient-aware institutional, commercial and outreach communication for public managers, operators, partners and stakeholders.
- Graphic materials — editable HTML/CSS/SVG/PDF-oriented institutional materials, with production constraints such as print sizes, bleed and export validation.
- Instagram and LinkedIn workflows — channel-specific social content, creative direction, CTAs and institutional tone for vBuz.
- Video strategy — audience-specific narrative, voice, scene planning and scripts for institutional and campaign video.
- CV tailoring — truthful, ATS-oriented CV adaptation with LaTeX and PDF outputs for a supplied role.
Connected tools
Notion
Connected through MCP for workspace search plus page and database workflows. This supports turning research, plans and project updates into shared documentation instead of isolated chat output.
Figma
Connected for design inspection, editable screen workflows, design-system-aware component work, diagrams and handoff between product design and implementation.
Visual production
ArtCraft and visual-production workflows are used alongside the design process for assets and presentation materials, with final review kept in the relevant design or production tool.
Working principles
- Use a focused workflow instead of a generic prompt when a repeatable process exists.
- Keep claims, external actions and sensitive operations under human review.
- Produce editable sources and validate the output in the environment where it will be used.
- Keep architecture and domain decisions in the application; AI accelerates execution and documentation, not accountability.
Type cd .. to return home.