PromptLikeEngineer vs dotfiles and Gists for prompts

• Zero cost, zero vendor. No account, no subscription, nobody who can change the pricing on you.
• It's just files. grep, git log and git diff all work. No new UI.
• Claude Code can read it directly. Reference a path and the prompt is in context.
• Full portability. It goes wherever git goes. There is never an export step.

If your collection is a handful of files you touch now and then, this is probably the right answer, and no tool should talk you out of it.

The costs don't show up on day one. They show up as the collection grows.

Search turns into grep plus memory. Fine at 10 files. At 100, you are remembering filenames or writing your own search script.

Git history is not per-prompt history. git log shows commits to the repo, not "every version of this one prompt, with a diff." You can get close with one prompt per file and careful commit messages, but that is discipline you maintain forever.

No {{variable}} filling. Placeholders mean find-and-replace by hand, or a script you now own.

Sync is only as good as your last git pull. Edit the same prompt on two machines and you find out at merge time.

Nothing an agent can query on its own. Claude Code can read a file you point it at, but it can't ask "which of my 80 prompts is about code review" unless you build that lookup. A folder doesn't expose a tool call.

• Search across titles and descriptions, plus tag filters.
• Per-prompt version history with diffs and rollback.
• Variable filling built into ple get --var and the MCP fill_variables tool.
• One library across CLI, web and MCP, without a manual pull.
• Agent-queryable: Claude Code can call search_prompts and get_prompt itself.

And it doesn't ask you to give up files. The CLI starts signed out and stores prompts as plain Markdown in ~/.promptlike/prompts/, so you can keep them in git. Sync to an account is one explicit command, and export stays open on every plan (JSON, an AGENTS.md section, or a Claude Skill).

npm i -g promptlike
ple save "code review" -m "Review this diff for bugs: {{diff}}"
ls ~/.promptlike/prompts/        # code-review.md, plain Markdown
ple sync --dry-run               # later: see what an upload would do

None of the hosted part is free forever at any size: Pro is $19/mo. The Free plan covers 25 prompts with up to 3 versions each, enough to try the workflow first.

You are not paying for storage; a folder is free storage. You are paying to not build and maintain the tooling around it: search, per-prompt history, variable filling, sync and agent access. If you enjoy that kind of work or already have scripts for it, the files win outright. If you would rather spend the time elsewhere, that is what you are buying.

Small collection, git is enough, no agent lookups   → dotfiles / Gist
Collection outgrew grep, Claude Code should find
prompts by name, or you test changes before keeping → PromptLikeEngineer

Comparing hosted tools instead? Read the best prompt manager for Claude Code or PromptLikeEngineer vs PromptHub.