GuideCodex4 min read
AGENTS.md for Codex: project instructions that last
Make repository conventions discoverable, verifiable and locally relevant so Codex can work more consistently.
Creagrid knowledge base
Insight for teams moving digital work forward.
Practical guides to AI, Codex, Claude Code, agents, automations, software and digital design — clearly explained and ready to apply.
Editorial picks
Editor’s selection
GuideClaude Code4 min read
CLAUDE.md: a useful project handbook for Claude Code
Give Claude Code the context every session needs without building a second internal wiki.
GuideCodex4 min read
A dependable Codex workflow from request to verification
Make agent work reviewable with a clear assignment, narrow change surface and evidence that the outcome works.
Topics
Go deeper on what is changing today
Follow a subject from first principles to a practical workflow through guides that build on one another.
AI in practice
From language models and governance to useful applications inside your team.
Codex
Work better with Codex, clear instructions and controllable workflows.
Claude Code
Practical patterns for context, tooling, planning and safe execution.
Agents & AGENTS.md
Build agent instructions that make teams, repositories and automations reliable.
New
1–9 of 15 articlesThe latest insights
AnalysisAI agents4 min read
Evaluating AI agents: measure the task, trajectory and risk
Build an evaluation set from real work and combine outcome, process evidence and human judgement.
GuideClaude Code4 min read
Claude Code in CI: useful pull request automation without a blank cheque
Restrict triggers, tokens and write scope so automated review remains a control rather than a new attack surface.
ChecklistClaude Code4 min read
Claude Code, data and privacy: a practical team review
Map data flows, retention and contractual choices before source code and business information enter an agent workflow.
GuideClaude Code4 min read
Claude Code hooks: fixed checks around flexible agent work
Use hooks for deterministic controls, not as a hidden collection of scripts nobody understands.
ChecklistClaude Code4 min read
Bound Claude Code safely with permissions and sandboxing
Combine approval rules, isolation, safe defaults and review so one mistake never receives broad reach.
GuideClaude Code4 min read
Subagents in Claude Code: specialise without creating chaos
Separate research, implementation and review only when task boundaries, permissions and hand-offs are clear.
GuideAI4 min read
Context engineering for AI agents: provide less, but better
Design what an agent sees, retains and retrieves at each moment instead of stacking all information into one prompt.
ExplainerAI agents4 min read
MCP explained: tools, resources and prompts without an open back door
Design AI connections as clear contracts with minimum permissions, explicit consent and traceable data flows.
ComparisonAI agents4 min read
Workflow or AI agent? Choose the simplest approach that works
Use fixed automation for predictable work and agent behaviour only where interpretation justifies added variability.