Extracting Workflow Data Efficiently

This skill teaches you how to extract specific data from Cromwell workflow metadata without loading the entire 100K+ token metadata blob into your context.

Recommended Usage Pattern

Step 1: Start with Summary Mode (Default)

get_job_metadata(workspace_namespace, workspace_name, submission_id, workflow_id)

Summary mode returns a structured, context-efficient summary (~1-2K tokens) including:

• •Workflow status and timing
• •Task counts by status
• •Failed task details with errors
• •Execution summary

This gives you the lay of the land without context exhaustion.

Step 2: Extract Specific Data with Extract Mode

Use semantic parameters to get exactly what you need:

Get a Specific Task Output

get_job_metadata(
    workspace_namespace, workspace_name, submission_id, workflow_id,
    mode="extract",
    task_name="illumina_demux",
    output_name="commonBarcodes"
)

Get Output from a Specific Shard

For scattered tasks with multiple shards:

get_job_metadata(
    workspace_namespace, workspace_name, submission_id, workflow_id,
    mode="extract",
    task_name="align_reads",
    shard_index=5,
    output_name="aligned_bam"
)

Get Runtime Attributes for All Tasks

Use wildcards with dot-path notation:

get_job_metadata(
    workspace_namespace, workspace_name, submission_id, workflow_id,
    mode="extract",
    field_path="calls.*.runtimeAttributes"
)

This returns a dictionary mapping task names to their runtime attributes.

Dot-Path Syntax Reference

The field_path parameter supports flexible extraction using dot notation:

Examples

# Get execution status of all tasks
field_path="calls.*.executionStatus"

# Get first shard's outputs for a specific task
field_path="calls.my_task[0].outputs"

# Get all runtime attributes
field_path="calls.*.runtimeAttributes"

# Get preemptible setting across all tasks
field_path="calls.*.runtimeAttributes.preemptible"

# Get backend status
field_path="calls.*[0].backendStatus"

Multiple Extractions

If you need multiple pieces of data, make multiple extract calls rather than trying to get everything at once. Each call returns only what you need:

# Call 1: Get specific output
get_job_metadata(..., mode="extract", task_name="task1", output_name="result")

# Call 2: Get runtime attributes
get_job_metadata(..., mode="extract", field_path="calls.task1[0].runtimeAttributes")

# Call 3: Get another task's output
get_job_metadata(..., mode="extract", task_name="task2", output_name="file")

This is more context-efficient than loading full metadata.

Error Handling

When extraction fails, the tool provides helpful error messages:

Task not found:

Task 'xyz' not found. Available tasks: ['workflow.task1', 'workflow.task2', ...]

Output not found:

Output 'xyz' not found for task 'task1'. Available outputs: ['file', 'count', 'summary']

Shard not found:

Shard 5 not found for task 'task1'. Available shards: [0, 1, 2, 3]

When to Use Each Mode

Anti-Patterns

Never do this:

• •Repeatedly call get_job_metadata without specifying what you need
• •Try to extract everything in one call

Do this instead:

• •Start with summary mode to understand the workflow
• •Make targeted extract calls for specific data you need
• •Use wildcards to batch-extract across tasks