lintlang 0.3.2

LintLang v0.3.2 — GitHub Action for AI agent and prompt linting

on Python PyPI

LintLang is a static prompt linter and agent-config linter for AI agent instructions, system prompts, and tool definitions. v0.3.2 adds first-party GitHub Action and pre-commit support for pull requests, CI, and local review.

The scan itself is offline and deterministic: it makes no LLM, API, telemetry, or network calls. LintLang flags ambiguous instructions, conflicting rules, missing execution limits, and schema-description mismatches before runtime testing.

What’s new

• First-party composite GitHub Action; FAIL findings block by default
• First-party pre-commit hook; verdict findings are advisory by default
• One-command uvx use and persistent isolated pipx installation
• GitHub Actions v7 maintenance

Add LintLang to GitHub Actions

steps:
  - uses: actions/checkout@v7
  - name: Lint agent instructions
    uses: hermes-labs-ai/lintlang@v0.3.2
    with:
      path: AGENTS.md

The Action blocks on FAIL (HIGH or CRITICAL) by default. Set fail-on: review to also block on REVIEW (MEDIUM) findings.

Try v0.3.2 locally

Run once with uvx—no persistent LintLang install:

uvx lintlang@0.3.2 scan AGENTS.md

Install LintLang as a persistent, isolated command:

pipx install lintlang
lintlang scan AGENTS.md

Already installed with pipx? Upgrade with:

pipx upgrade lintlang

Existing Action or pre-commit users can update their uses: or rev: pin to v0.3.2.

Verification and scope

No detector, verdict, or existing CLI behavior changed. At release commit 10a967d, hosted CI passed on Python 3.10–3.13; the Action accepted a clean fixture and rejected a FAIL fixture, and the pre-commit hook was exercised on Python 3.12.

LintLang performs static language checks. It does not evaluate runtime agent behavior, certify safety, or replace runtime evaluation or human review.

See the README for supported inputs and pre-commit setup, or the full changelog.