Step 3: Configure Thresholds¶

Now let's configure when validation should pass or fail based on your requirements.

Strict Mode¶

For production deployments, enable strict mode:

.github/workflows/validate-agent.yml

name: Validate A2A Agent

on:
  push:
    branches: [main]
  pull_request:
    branches: [main]

jobs:
  validate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: capiscio/validate-a2a@v1
        with:
          agent-card: './agent-card.json'
          strict: true

Strict mode:

Requires compliance score ≥ 95
Requires trust score ≥ 60 (if signatures present)
Requires availability score ≥ 80 (if tested)
Treats warnings as errors

Fail on Warnings¶

Even without strict mode, you can fail on any warnings:

- uses: capiscio/validate-a2a@v1
  with:
    agent-card: './agent-card.json'
    fail-on-warnings: true

Custom Score Thresholds¶

For custom thresholds, use the action outputs with a conditional step:

.github/workflows/validate-agent.yml

name: Validate A2A Agent

on:
  push:
    branches: [main]
  pull_request:
    branches: [main]

jobs:
  validate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Validate Agent Card
        id: validate
        uses: capiscio/validate-a2a@v1
        with:
          agent-card: './agent-card.json'

      - name: Check Compliance Score
        if: ${{ steps.validate.outputs.compliance-score < 80 }}
        run: |
          echo "::error::Compliance score (${{ steps.validate.outputs.compliance-score }}) is below threshold (80)"
          exit 1

      - name: Check Production Readiness
        if: ${{ steps.validate.outputs.production-ready == 'false' }}
        run: |
          echo "::warning::Agent is not production ready"
          # Don't fail, just warn

Different Thresholds for Different Branches¶

.github/workflows/validate-agent.yml

name: Validate A2A Agent

on:
  push:
    branches: [main, develop]
  pull_request:
    branches: [main, develop]

jobs:
  validate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      # Strict validation for main branch
      - name: Validate (Production)
        if: github.ref == 'refs/heads/main' || github.base_ref == 'main'
        uses: capiscio/validate-a2a@v1
        with:
          agent-card: './agent-card.json'
          strict: true

      # Lenient validation for develop branch
      - name: Validate (Development)
        if: github.ref == 'refs/heads/develop' || github.base_ref == 'develop'
        uses: capiscio/validate-a2a@v1
        with:
          agent-card: './agent-card.json'
          strict: false

Testing Live Endpoints¶

For staging/production validation, test the live endpoint:

- uses: capiscio/validate-a2a@v1
  with:
    agent-card: 'https://staging.myagent.com/.well-known/agent-card.json'
    test-live: true
    timeout: 30000  # 30 seconds for slower endpoints

Live Testing

Live testing sends actual A2A messages to your agent. Only enable this for deployed endpoints, not local files.

Skip Signature Verification¶

During development, you might not have signatures set up:

- uses: capiscio/validate-a2a@v1
  with:
    agent-card: './agent-card.json'
    skip-signature: true

Not for Production

Always verify signatures in production workflows. Unsigned agent cards have a trust score of 0.

Complete Example: Environment-Based Validation¶

.github/workflows/validate-agent.yml

name: Validate A2A Agent

on:
  push:
    branches: [main, staging, develop]
  pull_request:
    branches: [main, staging]

env:
  # Production thresholds
  PROD_COMPLIANCE_MIN: 95
  PROD_TRUST_MIN: 60
  # Staging thresholds
  STAGING_COMPLIANCE_MIN: 80
  STAGING_TRUST_MIN: 0

jobs:
  validate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Validate Agent Card
        id: validate
        uses: capiscio/validate-a2a@v1
        with:
          agent-card: './agent-card.json'

      - name: Enforce Production Thresholds
        if: github.ref == 'refs/heads/main'
        run: |
          compliance=${{ steps.validate.outputs.compliance-score }}
          trust=${{ steps.validate.outputs.trust-score }}

          if [ "$compliance" -lt "$PROD_COMPLIANCE_MIN" ]; then
            echo "::error::Compliance score ($compliance) below production minimum ($PROD_COMPLIANCE_MIN)"
            exit 1
          fi

          if [ "$trust" -lt "$PROD_TRUST_MIN" ]; then
            echo "::error::Trust score ($trust) below production minimum ($PROD_TRUST_MIN)"
            exit 1
          fi

          echo "✅ Production thresholds met!"

      - name: Enforce Staging Thresholds
        if: github.ref == 'refs/heads/staging'
        run: |
          compliance=${{ steps.validate.outputs.compliance-score }}

          if [ "$compliance" -lt "$STAGING_COMPLIANCE_MIN" ]; then
            echo "::error::Compliance score ($compliance) below staging minimum ($STAGING_COMPLIANCE_MIN)"
            exit 1
          fi

          echo "✅ Staging thresholds met!"

What's Next?¶

You now have:

Basic validation
Strict mode for production
Custom score thresholds
Environment-based validation

Let's add PR comments so reviewers can see results!

Back Continue