github-actions.md

GitHub Actions integration

MemoryGuard includes a small composite action for checking a project from CI. It installs the alpha directly from this repository, runs locally on the runner, and does not require a cloud account or an external LLM API.

Quick start

Create .github/workflows/memoryguard.yml in the project you want to inspect:

name: MemoryGuard

on:
  pull_request:

permissions:
  contents: read

jobs:
  doctor:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: atharvmantri/MemoryGuard@main
        with:
          project-path: .

The default mode is lenient: it fails on real installation/runtime errors but allows a project that has not created a .memoryguard store yet. This makes the first run useful as an installation check.

Once the project has an initialized store and you want warnings to fail CI, enable strict mode:

- uses: atharvmantri/MemoryGuard@main
  with:
    project-path: .
    strict: true

For production workflows, pin uses to a reviewed commit instead of main. The action uses uv and the repository's committed lockfile to run the source in the action checkout; it does not publish or install a package from PyPI. If the lockfile is stale, the action fails rather than silently resolving a new dependency graph in CI.

What the action does

• checks that the CLI and its local dependencies can start;
• inspects the selected project path;
• reports the current store/context-file state through memoryguard doctor;
• does not write project files or approve generated context.

The action is intentionally read-only. Use the local CLI's explicit memoryguard context update --yes command when you want to write generated context files as a separate, reviewed step.