Context Efficient Coding

A Codex skill that reduces unnecessary context, tool output, rereads, and response tokens without compromising correctness, completeness, or verification.

It is designed for implementation, debugging, refactoring, code review, testing, and repository exploration during long coding sessions.

Why

The skill prioritizes work quality over efficiency:

1. Correctness
2. Completeness
3. Verification
4. Context efficiency
5. Output efficiency

It saves context by narrowing investigation and retaining useful evidence—not by guessing, lowering reasoning effort, or skipping required tests.

Install

Requirements:

• Codex
• Python 3.10 or later for the optional helper scripts

macOS and Linux

mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills"
git clone https://github.com/andreynzz/context-efficient-coding.git \
  "${CODEX_HOME:-$HOME/.codex}/skills/context-efficient-coding"

Windows PowerShell

$codexHome = if ($env:CODEX_HOME) { $env:CODEX_HOME } else { Join-Path $HOME ".codex" }
New-Item -ItemType Directory -Force -Path (Join-Path $codexHome "skills")
git clone https://github.com/andreynzz/context-efficient-coding.git `
  (Join-Path $codexHome "skills\context-efficient-coding")

Start a new Codex task after installing so the skill can be discovered. It permits implicit invocation; invoke it explicitly with $context-efficient-coding when you want to make the intent clear.

How it works

The central inspection protocol is:

Search -> Slice -> Read -> Expand only if necessary

Before opening a large file, the skill searches for the relevant symbol, error, route, call site, or reference; reads the likely region; and expands only when the evidence requires it. It favors rg -n, narrow diffs, and focused tests.

For difficult debugging, migrations, cross-module refactors, and long tasks, it loads the dedicated heavy-work protocol. Benchmark guidance is loaded only when evaluating or tuning the skill.

Helper scripts

The scripts are optional, use only the Python standard library, make no network requests, and do not modify the target project. Resolve their path from the installed skill directory, then run them from the repository you are investigating.

Helper	Purpose	Example
repo_snapshot.py	Summarizes the repository, working tree, project markers, and active AGENTS.md chain without dumping file contents.	python3 <skill-dir>/scripts/repo_snapshot.py
smart_read.py	Prints numbered file regions or regex-match context with explicit truncation.	python3 <skill-dir>/scripts/smart_read.py src/app.py --match 'createUser' --context 30
compact_run.py	Retains a noisy command's full log; previews the tail on success and the head plus tail on failure.	python3 <skill-dir>/scripts/compact_run.py -- npm test

On Windows, use an available Python 3 command such as py -3 instead of python3 when appropriate.

Guidance included

• Heavy-work protocol for difficult debugging, migrations, architecture work, and cross-module refactors.
• Benchmark protocol for LIGHT, NORMAL, and HEAVY A/B evaluations. Optimizations are rejected if they reduce quality, verification, or total time to a correct outcome.

Test

Python 3.10 remains the minimum because the helpers use modern type-union syntax without compatibility shims. Run the same compile check and standard-library test suite used by CI:

python3 -m py_compile scripts/repo_snapshot.py scripts/smart_read.py scripts/compact_run.py
python3 -m unittest discover -s tests -v

GitHub Actions runs these checks on pushes and pull requests targeting main. Use the script help output to see all options:

python3 scripts/smart_read.py --help
python3 scripts/compact_run.py --help

Contributing

Keep the entrypoint focused: universal rules belong in SKILL.md; conditional workflows belong in references/; deterministic, reusable behavior belongs in scripts/. Test every change to a helper script and avoid adding dependencies unless they provide a concrete benefit.

License

Distributed under the MIT License.