plugin-tests.yml

# Consolidated CI for the stable e2a plugin and experimental e2a Labs plugin.
# Package validation and skill suites run in parallel on every pull request and
# main push so the release gate cannot be skipped by GitHub's path-filter file
# limit. The model-backed Agentify fixtures run only when their inputs change,
# and skip cleanly when model credentials are absent.
name: Plugin tests

on:
  push:
    branches: [main]
  pull_request:
  workflow_dispatch: {}

permissions:
  contents: read

concurrency:
  group: ${{ github.workflow }}-${{ github.event_name == 'pull_request' && github.ref || github.sha }}
  cancel-in-progress: ${{ github.event_name == 'pull_request' }}

jobs:
  package:
    name: Package and manifests
    runs-on: ubuntu-latest
    timeout-minutes: 10
    env:
      CLAUDE_CODE_VERSION: "2.1.226"
    steps:
      - uses: actions/checkout@v7
        with:
          fetch-depth: 0
      - uses: actions/setup-node@v7
        with:
          node-version: "22"
      # The version-bump gate diffs the base against the working tree, so it
      # must run before anything that mutates the checkout. In particular
      # `npm ci` inside the vendored email-evals runtime creates ignored local
      # node_modules on Linux (including platform-specific esbuild binaries),
      # which should not affect the version-bump diff.
      - name: Require plugin version bumps in pull requests
        if: github.event_name == 'pull_request'
        env:
          PLUGIN_VERSION_BASE: ${{ github.event.pull_request.base.sha }}
        run: node scripts/check-plugin-version-bump.mjs "$PLUGIN_VERSION_BASE"
      - name: Require plugin version bumps pushed to main
        if: github.event_name == 'push' && github.event.before != '0000000000000000000000000000000000000000'
        env:
          PLUGIN_VERSION_BASE: ${{ github.event.before }}
        run: node scripts/check-plugin-version-bump.mjs "$PLUGIN_VERSION_BASE"
      - run: npm install --global "@anthropic-ai/claude-code@${CLAUDE_CODE_VERSION}"
      - name: Install email-evals runtime dependencies
        run: npm ci --ignore-scripts --prefix plugins/e2a/skills/email-evals/runtime
      - name: Run email-evals runtime tests
        run: npm test --prefix plugins/e2a/skills/email-evals/runtime
      - name: Run plugin contract tests
        run: node --test scripts/plugin-agent-guidance.test.mjs scripts/plugin-core-skills.test.mjs scripts/plugin-packaging.test.mjs scripts/plugin-email-evals.test.mjs scripts/check-plugin-version-bump.test.mjs scripts/validate-plugin.test.mjs
      - name: Validate repository plugin contracts
        run: node scripts/validate-plugin.mjs
      - name: Validate Claude plugin manifests
        run: |
          claude plugin validate --strict plugins/e2a
          claude plugin validate --strict plugins/e2a-labs
  agentify:
    name: Agentify deterministic suite
    runs-on: ubuntu-latest
    timeout-minutes: 10
    steps:
      - uses: actions/checkout@v7
      - uses: actions/setup-node@v7
        with:
          node-version: "22"
      - uses: actions/setup-python@v7
        with:
          python-version: "3.12"
      - name: Install PyYAML for config validation
        run: pip install --quiet pyyaml
      - name: Run Agentify tests
        run: bash plugins/e2a-labs/skills/agentify/test/run.sh

  autopilot:
    name: Autopilot suite
    runs-on: ubuntu-latest
    timeout-minutes: 10
    steps:
      - uses: actions/checkout@v7
      - uses: actions/setup-node@v7
        with:
          node-version: "22"
      - name: Run Autopilot tests
        run: node --test plugins/e2a-labs/skills/autopilot/test/*.test.mjs

  tether:
    name: Tether self-test
    runs-on: ubuntu-latest
    timeout-minutes: 10
    steps:
      - uses: actions/checkout@v7
      - uses: actions/setup-node@v7
        with:
          node-version: "22"
      - uses: actions/setup-python@v7
        with:
          python-version: "3.12"
      - name: Run Tether runtime self-test
        run: bash plugins/e2a-labs/skills/tether/tether.sh _selftest

  agentify-fixtures:
    name: Agentify live fixtures
    if: github.event_name == 'pull_request' || github.event_name == 'workflow_dispatch'
    runs-on: ubuntu-latest
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@v7
        with:
          fetch-depth: 0
      - name: Detect fixture-relevant changes
        id: fixture-changes
        env:
          EVENT_NAME: ${{ github.event_name }}
          BASE_SHA: ${{ github.event.pull_request.base.sha }}
        run: |
          if [ "$EVENT_NAME" = "workflow_dispatch" ]; then
            run_fixtures=true
          elif git diff --quiet "$BASE_SHA"...HEAD -- \
            plugins/e2a-labs/skills/agentify/templates/runtime-skill \
            plugins/e2a-labs/skills/agentify/templates/workflows/feedback-triage.yml.tmpl \
            plugins/e2a-labs/skills/agentify/examples/e2a/autonomous-repo.config.yml \
            plugins/e2a-labs/skills/agentify/test/fixtures \
            .github/workflows/plugin-tests.yml; then
            run_fixtures=false
          else
            run_fixtures=true
          fi
          echo "run_fixtures=$run_fixtures" >> "$GITHUB_OUTPUT"
      - uses: actions/setup-node@v7
        if: steps.fixture-changes.outputs.run_fixtures == 'true'
        with:
          node-version: "22"
      - name: Install Claude Code and fixture dependencies
        if: steps.fixture-changes.outputs.run_fixtures == 'true'
        run: |
          npm install --global @anthropic-ai/claude-code
          npm install --prefix plugins/e2a-labs/skills/agentify/test/fixtures/harness
      - name: Run Agentify lane fixtures
        if: steps.fixture-changes.outputs.run_fixtures == 'true'
        env:
          CLAUDE_CODE_OAUTH_TOKEN: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
        run: bash plugins/e2a-labs/skills/agentify/test/fixtures/run-fixtures.sh