Changelog

All notable changes to PromptLint are documented here.

v1.4.0

Released: April 2026

New Rule

• output-length-missing — INFO-level completeness rule that fires when a prompt contains a task verb (write, generate, list, etc.) but no length constraint (word count, sentence limit, brevity directive). Total: 21 rules, 5 auto-fixable.

New CLI Features

Flag	Description
--show-score	Display health score (0–100) + letter grade (A–F) and per-category breakdown
--badge	Output a Shields.io badge URL + markdown snippet for the health score
--compare FILE_A FILE_B	Lint two prompts and show score delta per category — useful for validating rewrites
--update-baseline	Write current findings to .promptlintbaseline; future runs suppress known issues
--install-hooks	Install a pre-commit git hook that lints staged .txt/.md/.prompt files

Improvements

• Message array input — both CLIs now accept OpenAI/Anthropic [{role, content}…] JSON arrays via --text or stdin; content fields are joined and linted as a single prompt
• TS autofix parity — verbosity-redundancy auto-fix expanded from 7 to 41 replacement patterns to match the Python implementation
• Docs site redesigned to match the main landing page (black/green terminal aesthetic, JetBrains Mono, film grain overlay)
• New Config Examples page with 10 ready-to-copy .promptlintrc starters for common use cases

---

v1.3.0

Released: 2024

New Rules

• jailbreak-pattern — Detects jailbreak attempts and persona hijacking
• secret-in-prompt — Detects hardcoded API keys, passwords, connection strings
• pii-in-prompt — Detects emails, phone numbers, SSNs, credit card numbers
• context-injection-boundary — Enforces trust boundaries on user-injected content
• role-clarity — Checks for missing role/persona definitions
• output-format-missing — Checks for missing output format specification
• hallucination-risk — Detects patterns that increase hallucination likelihood

Improvements

• SARIF v2.1.0 output format for GitHub Security tab integration
• VS Code extension with real-time linting and quick-fix support
• GitHub Action with PR annotations and SARIF upload
• Python/TypeScript parity tests in CI
• Score (0–100) and grade (A–F) in all output formats

Rule Count

Total: 20 rules, 5 auto-fixable

---

v1.2.0

New Features

• --show-dashboard flag for cost savings visualization
• Monthly and annual cost projection in dashboard
• --rule flag to run a single rule
• --explain flag for rule documentation in the terminal

---

v1.1.0

New Features

• --format sarif output (v2.1.0)
• Score and grade calculation
• calls_per_day config for cost projection

New Rules

• completeness-edge-cases
• consistency-terminology

---

v1.0.0

Initial Stable Release

• 13 rules across security, cost, quality, and structure
• --fix auto-fix with 5 fixable issues
• --format json for CI/CD
• .promptlintrc YAML configuration
• Python CLI (pip) and TypeScript CLI (npm)
• GitHub Action (AryaanSheth/promptlint@v1)

Rules at Launch

• cost, cost-limit
• prompt-injection
• structure-sections
• clarity-vague-terms
• specificity-examples, specificity-constraints
• politeness-bloat
• verbosity-redundancy, verbosity-sentence-length
• actionability-weak-verbs
• consistency-terminology, completeness-edge-cases

---

v0.2.0

• Structure flexibility: XML, headings, and Markdown all recognized
• Politeness configuration: allow_politeness: true demotes to INFO
• Improved auto-fix: structure scaffold respects structure_style config

v0.1.0

• Initial release with 13 rules and 5 auto-fixes