Cheatsheets as AI Context

Why format matters for AI

When you paste a cheatsheet into Claude or include it via @, the AI reads it as plain text. A well-formatted cheatsheet gives the AI structured, unambiguous information. A poorly formatted one wastes context and produces worse output.

Use fenced code blocks for commands

AI tools treat fenced code blocks as literal, executable content. Inline code in prose gets mixed up with the surrounding text.

git stash        # good: in a code block, clearly a command

Avoid: “you can use git stash to save your changes” — the command is buried in prose.

Use consistent comment style

Pick one comment style and use it everywhere. For shell commands, # inline is standard:

git commit -m "message"    # commit staged changes
git push                   # push to remote

For non-shell content (prompts, shortcuts), a # or → arrow works:

Ctrl + Z    # undo
Ctrl + Y    # redo

Keep sections flat

Deeply nested structure confuses AI context parsing. Two levels max:

## Section heading
command    # description
command    # description

Avoid: ### Sub-sub-section — the AI loses track of which group a command belongs to.

Label each section clearly

Section headings are how the AI knows which commands belong to which task. Make them task-oriented:

## Start a Session      ✅ task-oriented
## Commands             ❌ too vague

Include the “when to use” in the comment

The inline comment should say why, not just what:

git restore .    # discard ALL unstaged changes — destructive, can't undo

This lets the AI reason about whether a command is appropriate for a given situation, not just what it does syntactically.

Keep it short

The longer a cheatsheet, the more context it consumes. A 40-line cheatsheet included in a prompt uses roughly the same tokens as a paragraph of instructions — worth it. A 200-line cheatsheet starts competing with the actual task for context space.