Dataset Validator - Pre-Deployment (Example Skill Hook)

<CRITICAL_RULES> IMPORTANT: Skill Hook Interface

Receive WorkflowContext via environment variables (set by invoke-skill-hook.sh)
Return structured WorkflowResult as JSON to stdout
Exit with code 0 for success, 1 for failure
Always include success, messages, errors in result
WorkflowResult must be valid JSON </CRITICAL_RULES>

Optional: WorkflowContext JSON file passed as argument </INPUTS>

<COMPLETION_CRITERIA> ✅ WorkflowContext read and parsed ✅ Dataset validation checks executed ✅ Validation results documented ✅ WorkflowResult JSON output to stdout ✅ Appropriate exit code returned </COMPLETION_CRITERIA>

<ERROR_HANDLING> If validation fails:

Set success: false
Populate errors array with specific issues
Include suggestions in messages
Exit with code 1
Skill hook execution will be blocked (if required: true) </ERROR_HANDLING>

<EXAMPLE_EXECUTION>

bash

# Environment variables set by faber-cloud hook executor
export FABER_CLOUD_ENV="test"
export FABER_CLOUD_OPERATION="deploy"
export FABER_CLOUD_HOOK_TYPE="pre-deploy"
export FABER_CLOUD_PROJECT="myproject"
export FABER_CLOUD_TERRAFORM_DIR="./infrastructure/terraform"

# Skill invoked by invoke-skill-hook.sh
/skill dataset-validator-deploy-pre

# Expected output (JSON):
{
  "success": true,
  "messages": ["Validation passed"],
  "warnings": [],
  "errors": [],
  "artifacts": {},
  "executionTime": 1234,
  "timestamp": "2025-11-07T12:00:00Z",
  "skillName": "dataset-validator-deploy-pre"
}

</EXAMPLE_EXECUTION>

Implementation

Step 1: Read WorkflowContext

bash

# Read context from environment variables
ENV=${FABER_CLOUD_ENV:-unknown}
OPERATION=${FABER_CLOUD_OPERATION:-unknown}
HOOK_TYPE=${FABER_CLOUD_HOOK_TYPE:-unknown}
PROJECT=${FABER_CLOUD_PROJECT:-unknown}

echo "🎯 STARTING: Dataset Validator (Pre-Deployment)"
echo "Environment: $ENV"
echo "Operation: $OPERATION"
echo "Hook Type: $HOOK_TYPE"
echo "Project: $PROJECT"
echo "───────────────────────────────────────"

Step 2: Validate Datasets

bash

# Example validation logic
DATASET_DIR="./datasets"
ISSUES_FOUND=0
MESSAGES=()
WARNINGS=()
ERRORS=()

# Check datasets exist
if [ ! -d "$DATASET_DIR" ]; then
  ERRORS+=("Dataset directory not found: $DATASET_DIR")
  ISSUES_FOUND=$((ISSUES_FOUND + 1))
else
  MESSAGES+=("Dataset directory found: $DATASET_DIR")

  # Count datasets
  DATASET_COUNT=$(ls -1 "$DATASET_DIR"/*.csv 2>/dev/null | wc -l)
  MESSAGES+=("Found $DATASET_COUNT dataset files")

  # Validate each dataset
  for dataset in "$DATASET_DIR"/*.csv; do
    if [ -f "$dataset" ]; then
      # Check file size
      SIZE=$(stat -f%z "$dataset" 2>/dev/null || stat -c%s "$dataset" 2>/dev/null)
      if [ $SIZE -gt 10000000 ]; then
        WARNINGS+=("Dataset $(basename $dataset) is large: ${SIZE} bytes")
      fi

      # Validate CSV structure (example: check header row)
      if ! head -n1 "$dataset" | grep -q ","; then
        ERRORS+=("Dataset $(basename $dataset) appears invalid (no CSV header)")
        ISSUES_FOUND=$((ISSUES_FOUND + 1))
      fi
    fi
  done
fi

Step 3: Generate WorkflowResult

bash

# Build result JSON
START_TIME=$(date +%s)
END_TIME=$(date +%s)
EXECUTION_TIME=$((END_TIME - START_TIME))
TIMESTAMP=$(date -u +"%Y-%m-%dT%H:%M:%SZ")

SUCCESS="true"
if [ $ISSUES_FOUND -gt 0 ]; then
  SUCCESS="false"
fi

# Convert arrays to JSON
MESSAGES_JSON=$(printf '%s\n' "${MESSAGES[@]}" | jq -R . | jq -s .)
WARNINGS_JSON=$(printf '%s\n' "${WARNINGS[@]}" | jq -R . | jq -s .)
ERRORS_JSON=$(printf '%s\n' "${ERRORS[@]}" | jq -R . | jq -s .)

# Output WorkflowResult as JSON
cat <<EOF
{
  "success": $SUCCESS,
  "messages": $MESSAGES_JSON,
  "warnings": $WARNINGS_JSON,
  "errors": $ERRORS_JSON,
  "artifacts": {
    "datasetsValidated": $DATASET_COUNT,
    "issuesFound": $ISSUES_FOUND
  },
  "executionTime": $EXECUTION_TIME,
  "timestamp": "$TIMESTAMP",
  "skillName": "dataset-validator-deploy-pre"
}
EOF

# Exit with appropriate code
if [ $ISSUES_FOUND -gt 0 ]; then
  exit 1
else
  exit 0
fi

Testing

Test independently:

bash

# Set up test environment
export FABER_CLOUD_ENV="test"
export FABER_CLOUD_OPERATION="deploy"
export FABER_CLOUD_HOOK_TYPE="pre-deploy"
export FABER_CLOUD_PROJECT="test-project"

# Create test datasets
mkdir -p ./datasets
echo "id,name,value" > ./datasets/test1.csv
echo "1,test,100" >> ./datasets/test1.csv

# Invoke skill
/skill dataset-validator-deploy-pre

# Expected: Success with messages about datasets found

Test as hook:

json

{
  "hooks": {
    "pre-deploy": [
      {
        "type": "skill",
        "name": "dataset-validator-deploy-pre",
        "required": true,
        "failureMode": "stop",
        "timeout": 300
      }
    ]
  }
}

bash

# Run hook executor
bash plugins/faber-cloud/skills/cloud-common/scripts/execute-hooks.sh pre-deploy test ./infrastructure/terraform

# Expected: Hook executes skill and reports results

Installation

To use this skill hook in your project:

Copy skill to project:

bash

mkdir -p .claude/skills/dataset-validator-deploy-pre
cp dataset-validator-deploy-pre-SKILL.md .claude/skills/dataset-validator-deploy-pre/SKILL.md

Configure hook in faber-cloud.json:

json

{
  "hooks": {
    "pre-deploy": [
      {
        "type": "skill",
        "name": "dataset-validator-deploy-pre",
        "required": true,
        "failureMode": "stop",
        "timeout": 300,
        "description": "Validate datasets before deployment"
      }
    ]
  }
}

Test the hook:

bash

# Test skill independently
/skill dataset-validator-deploy-pre

# Test full deployment workflow
/fractary-faber-cloud:deploy-apply --env=test

Benefits

✅ Reusable - Share across projects ✅ Testable - Test independently with /skill ✅ Discoverable - Visible in /help ✅ Structured - Type-safe interfaces ✅ Maintainable - Clear documentation ✅ Extensible - Easy to customize

See Also:

Search AI Tools

dataset-validator-deploy-pre

Install this agent skill to your Project

SKILL.md

Dataset Validator - Pre-Deployment (Example Skill Hook)

Implementation

Testing

Installation

Benefits