Prompt Library coding beginner

ChatGPT Prompts for Code Comments: Documentation That Adds Value

ChatGPT prompts for writing code comments and documentation. Inline comments, JSDoc/docstrings, and README sections that help future developers.

Tested on: GPT-4oClaude 4Gemini 2.5

The Prompt

Act as a senior engineer who is known for writing exceptionally clear code documentation.
Write documentation for:
Code: [PASTE FUNCTION/CLASS/MODULE]
Language: {language — affects docstring format}
Docstring format: {JSDoc / Google / NumPy / reStructuredText / plain}
Audience: {junior engineers / mid-level / senior / external API consumers}
Documentation level: {inline comments only / function/class docstrings / full module documentation}

Output:
1. Function/class docstring (complete — includes description, parameters, return value, raises, example)
2. Inline comments (for complex logic only — added to the code with comments, not as a list)
3. Module-level documentation (if documenting a full file — purpose, usage, dependencies, author notes)
4. README section (if this is a public API — installation + quickstart code example)
5. What NOT to comment (identify any redundant or obvious comments in the original code)

Comment quality rules:
- Comments explain WHY, not WHAT (the code shows what it does — comments explain why it does it)
- No comments that merely translate code to English ("increment i by 1" over i++)
- Complex algorithms get a reference to the algorithm name or source
- TODO comments must include: who, what, and by when (or a ticket reference)
- Docstrings must include at least one working code example

Constraints:
- Do not add comments to self-explanatory code — over-commenting reduces readability
- Every parameter in a docstring must include its type and constraints (e.g. must be > 0)
- Code examples in docstrings must be runnable, not pseudocode
- Identify and flag any areas of the code where behavior is non-obvious and documentation is critical

Variables to fill in

  • {code} The function, class, or module to document
  • {docstring format} JSDoc, Google, NumPy, reStructuredText, or plain
  • {audience} Who will read this documentation — affects technical depth
  • {documentation level} Inline comments, docstrings, or full module documentation

How to use this prompt

  1. Paste legacy code into this prompt to retrofit documentation before a refactoring sprint
  2. Run this on any public API function before merging to main
  3. Use the 'what NOT to comment' section to clean up over-commented legacy code
  4. Have new team members add docstrings to code they just shipped — forces deep reading of what they wrote
Code editor with well-commented function showing JSDoc style documentation
Photo by Luca Bravo on Unsplash

Good comments explain why, not what

The most common bad comment is one that translates code to English: // increment the counter above count++. This adds noise without adding information — anyone who can read the code already knows what it does. The comment that adds value is the one that explains why: // Counter must be incremented before the callback fires to prevent race condition in IE11.

Docstrings are your API’s user manual

For any function that other developers will call — especially public APIs — the docstring is the first (and often only) documentation a developer reads. A complete docstring that includes parameter types, return value, raises, and a working example eliminates the most common integration questions before they become support tickets.

Over-commenting is its own form of technical debt

Codebases with comments on every line become harder to read, not easier. When code changes but comments don’t get updated — which is always — you end up with misleading documentation that actively misleads readers. The prompt’s ‘what NOT to comment’ section specifically identifies obvious comments to remove, keeping the documentation signal-to-noise ratio high.