Prepare a project for Claude Code and Codex: from setup to the first change
From an empty folder to a verified first change: AGENTS.md and CLAUDE.md instructions, tests, permissions and a handoff for coding agents.
Blog / Matteo Vincenti
Notes from work.
I write about AI in everyday work, software development and how teams work together. I start with business problems and explore the technical decisions.
20 articles, page 1 of 2
“Act as an expert”: what a role prompt changes
What does a role add to a prompt? Compare persona, message roles and explicit criteria with four complete prompts for the same caching decision.
UUIDv7 or bigint in PostgreSQL: arrival order matters too
A Docker comparison of bigint, UUIDv4 and UUIDv7: insert time, index size, WAL and arrival order. Includes measured results and runnable code.
Debugging with an AI agent: challenge the diagnosis before fixing the code
Reproduce a JavaScript date bug in Docker, distinguish two diagnoses and preserve regression tests before asking an AI agent to patch the code.
RAG, CAG and KAG: choosing how AI accesses knowledge
Compare RAG, CAG and KAG through documents, caching and relationships. Practical decision criteria, update costs and open-source projects.
Prompt injection: what can leave an agent’s environment
Follow a hostile proposal through a data transfer. A Docker lab separates file access, network reachability, recipients and legitimate work.
Enterprise RAG: why it retrieves the wrong document
A reproducible lab compares full-text, embeddings and hybrid search to show when to fix missing sources, eligibility or ranking.
MCP and skills: measure and reduce tokens without losing context
A reproducible MCP lab measures local token counts, tests filters and projections, and checks that shorter results preserve the evidence needed.
Harness engineering: which controls an agent actually needs
A Docker lab shows how an agent harness controls permissions, retries, budgets and evidence, and when to choose an existing runtime.
Code review with AI agents: when a pull request is ready
A small pull request can pass its tests and still be wrong. A Docker example shows how to expose the defect, fix it and explain the acceptance decision.
Stateful agents: recovering from a crash without duplicating actions
A Docker lab with LangGraph shows what runs again after a crash and how checkpoints and idempotency keys prevent duplicate tickets.
Meta repositories and AI agents: a project map or pinned revisions?
Decide when a project map is enough and when to pin revisions. A complete submodule setup with manifests, agent instructions and a runnable Git example.
Contact