A Codex skill for writing and maintaining execution plans: self-contained Markdown files that keep longer coding tasks legible while an agent researches, edits, tests, and changes course.
This repository packages the execplans skill directly under .codex/skills. There is no symlink,
agent profile, or repository-specific wrapper.
ExecPlans are a practical convention I use when a task is too large for "just patch the code." The plan names the goal, records progress and decisions, captures surprises, and stays on disk so either a person or an agent can resume from the file instead of relying on chat history.
For more context, see Using PLANS.md for multi-hour problem solving in the OpenAI Cookbook and my OpenAI DevDay talk, Shipping with Codex.
Clone this repository into a project root:
git clone https://github.com/friel-openai/plans.md.git
cd plans.md
Or copy the skill directory into another Codex configuration:
.codex/skills/execplans
Ask Codex to use the execplans skill for work that needs a durable plan:
Use the execplans skill. Create an ExecPlan for this migration, keep it updated while you work,
and implement the plan end to end.
Codex will read the bundled .codex/skills/execplans/references/PLANS.md specification, create or
refresh an ExecPlan as a Markdown file on disk, and keep its living sections current as work
proceeds.
ExecPlans are most useful for multi-hour implementations, large refactors, migrations, debugging investigations, research-heavy changes, and any task where future-you should be able to reopen one file and understand what happened.