Codex CLI Intermediate
Push Codex past single commands. Write structured AGENTS.md files, compose multi-step pipelines, recover from tool-call failures, and tune sandbox permissions.
LAST UPDATED · MAY 5, 2026 · BUILDER COHORT OPENS MONTHLY
§ 01 · OUTCOMES
What you'll ship
Not lessons watched. Artifacts you keep. Every outcome is work you can defend in review and reuse on Monday.
Structured AGENTS.md authoring
Leave with hands-on reps in structured agents.md authoring — applied to your own work, not toy exercises.
Multi-step task pipelines
Leave with hands-on reps in multi-step task pipelines — applied to your own work, not toy exercises.
Tool call failure handling
Leave with hands-on reps in tool call failure handling — applied to your own work, not toy exercises.
§ 02 · OVERVIEW
The vibe coding workflow
This course teaches the vibe coding workflow behind Codex CLI Intermediate: spec-driven prompting, diff discipline, and review-before-merge. You write the spec, the agent writes the diff, and you defend the work in review.
Every week ends with one shipped artifact — a CLAUDE.md revision, a reviewed PR, a context budget you can keep. No theory, no toy exercises. You leave with proof of work.
§ 03 · CURRICULUM
Six weeks, six artifacts
WEEK 01
Structured AGENTS.md authoring
Practice structured agents.md authoring against a real repo with live diff review.
SHIPS · A REVIEWED ARTIFACT
SHIPSA REVIEWED ARTIFACT
WEEK 02
Multi-step task pipelines
Practice multi-step task pipelines against a real repo with live diff review.
SHIPS · A REVIEWED ARTIFACT
SHIPSA REVIEWED ARTIFACT
WEEK 03
Tool call failure handling
Practice tool call failure handling against a real repo with live diff review.
SHIPS · A REVIEWED ARTIFACT
SHIPSA REVIEWED ARTIFACT
WEEK 04
Sandbox permission tuning
Practice sandbox permission tuning against a real repo with live diff review.
SHIPS · A REVIEWED ARTIFACT
SHIPSA REVIEWED ARTIFACT
§ 04 · COMPETENCIES
What you'll leave knowing
- Structured AGENTS.md authoring
- Multi-step task pipelines
- Tool call failure handling
- Sandbox permission tuning
§ 06 · CONTINUE
Pair it with
Claude Code Beginners
Run Claude Code as a terminal-first agent. Write CLAUDE.md, scope context with @-files, and defend every multi-file diff before it merges.
Claude Code Intermediate
Push Claude Code past single-file tasks. Chain multi-step sessions, write slash commands, manage large context windows, and wire Claude into CI.
Claude Code Advanced
Run Claude Code at production scale. Build CLAUDE.md systems, enforce review gates across teams, and instrument every session for audit and compliance.