~/tools / prompt-linter

Prompt linter

Prompts fail for boring, repeatable reasons: a vague 'etc.', no output format, untrusted input without delimiters, 'concise' and 'comprehensive' in the same breath. This linter runs deterministic checks against those patterns — compiler-style findings with a score — before the failure reaches production.

TOOL 07

Prompt linter

ESLint for the sentences you ship to production

paste a system prompt — lints as you type

// findings appear here, compiler-style

Heuristics from real incident reviews: vague terms, missing output formats, absent delimiters and conflicting absolutes cause most prompt failures. Deterministic checks — no model, no API, no judgment calls.

tools.siblings

Token & cost metercount tokens · compare API cost across models RAG capacity plannercorpus size → vectors, RAM, embed cost Provider request translatorOpenAI ⇄ Anthropic/Bedrock request JSON GPU VRAM calculatorwill that model fit on your card?JSONL dataset inspectorfine-tune data sanity check, fully local Prompt diffword-level diff for prompt iteration