Validation Pipeline

Validation Pipeline

Purpose

Run structured validation pipelines that execute steps sequentially, enforce gates at blocking failures, and produce scored reports. Uses the src/validation/pipeline.ts framework with 13 requirements validation steps (BMAD-003).

Activation

• When validating requirements documents
• When running structured quality gates
• When assessing document completeness, testability, or traceability
• When invoked via /validation-pipeline

Quick Start

# Validate a requirements document (all 13 steps)
/validation-pipeline requirements docs/requirements.md

# Validate with specific steps only
/validation-pipeline requirements docs/requirements.md --steps format-check,completeness-check,invest-criteria

# Continue past blocking failures
/validation-pipeline requirements docs/requirements.md --continue-on-failure

# Output as JSON
/validation-pipeline requirements docs/requirements.md --json

Workflow

Step 1: Read the Target Document

Read the file specified by the user. If no file is provided, ask for one.

Read the target document using the Read tool.
Store the content for pipeline execution.

Step 2: Select Pipeline

Choose the appropriate pipeline based on the user's request:

Additional pipelines can be created by defining new step sets in src/validation/steps/.

Step 3: Execute Pipeline

The pipeline framework (src/validation/pipeline.ts) handles execution:

1. Sequential execution — steps run in order, each receiving results from prior steps
2. Gate enforcement — blocking steps that fail halt the pipeline (unless --continue-on-failure)
3. Per-step scoring — each step produces a 0-100 score with findings and evidence
4. Weighted rollup — overall score uses category weights (format=10%, content=30%, quality=25%, traceability=20%, compliance=15%)

Requirements Pipeline Steps (13 total)

Step 4: Report Results

Format the pipeline result as a structured report:

# Validation Report: Requirements Pipeline

**Overall**: PASS/FAIL/WARN | **Score**: 85/100 | **Duration**: 42ms

## Step Results
| # | Step | Status | Score | Findings | Duration |
|---|------|--------|-------|----------|----------|
| 1 | Format Check | PASS | 100 | 0 | 2ms |
| 2 | Completeness | WARN | 60 | 2 | 5ms |
...

## Blockers
- (blocking findings listed here)

## All Findings
- [HIGH] Missing acceptance criteria: Requirement US-104 has no AC
- [MEDIUM] Vague term: "should" used 5 times without specifics
...

Step 5: Record Learning

After pipeline execution, record the outcome for learning:

// Store validation pattern
memory store --namespace validation-pipeline --key "req-validation-{timestamp}" --value "{score, findings_count, halted}"

Parameters

Integration Points

• qe-requirements-validator agent — delegates structured validation to this pipeline
• qe-quality-gate agent — uses pipeline for gate evaluation
• YAML Pipelines — can invoke validation steps as workflow actions
• MCP — accessible via pipeline_validate tool

Output Schema

The pipeline produces a PipelineResult object (see schemas/output.json):

{
  pipelineId: string;
  pipelineName: string;
  overall: 'pass' | 'fail' | 'warn';
  score: number;           // 0-100 weighted average
  steps: StepResult[];     // per-step details
  blockers: Finding[];     // blocking findings
  halted: boolean;
  haltedAt?: string;       // step ID where halted
  totalDuration: number;
  timestamp: string;
}

Error Handling

• Step throws exception — captured as a FAIL with critical finding, pipeline continues or halts per severity
• File not found — report error, do not run pipeline
• Empty document — format-check step will catch this as a blocking failure