sdlc-outcome-validator

# SDLC Outcome Validator

## Overview

Provides a reusable validation loop pattern that all SDLC agents use as their final step before delivering output. Instead of each agent independently implementing success metrics checking and loop-back logic, agents invoke this skill's standardized pattern.

## When to Use This Skill

- As the final step of any SDLC agent workflow before output delivery
- When an agent needs to validate its output against defined success criteria
- When loop-back to earlier steps is needed because validation fails

## Unitary Function

Given an agent's success metrics and its current output, evaluate each metric, report pass/fail status, and determine whether the output is ready for delivery or needs to loop back to a prior step.

## Validation Loop Pattern

Every SDLC agent MUST execute this pattern as its final workflow step:

### Step 1: Collect Success Metrics

Read the agent's `## Success Metrics` section. Each bullet point is a discrete criterion to validate.

### Step 2: Evaluate Each Metric

For each success metric, perform a concrete validation:

```
METRIC: "<success criterion text>"
CHECK:  <what was actually checked>
RESULT: PASS | FAIL | PARTIAL
EVIDENCE: <specific evidence supporting the result>
REMEDIATION: <if FAIL/PARTIAL -- which workflow step to return to and what to fix>
```

### Step 3: Determine Outcome

| Result | Action |
|---|---|
| All PASS | Proceed to output delivery |
| Any FAIL | Loop back to the step identified in REMEDIATION. Re-execute that step. Return to validation. Maximum 2 loop-back iterations. |
| PARTIAL only (no FAIL) | Proceed to output delivery with PARTIAL items noted as caveats in the output |

### Step 4: Record Validation Result

Include validation status in the output JSON:

```json
"outcome_validation": {
  "validator": "sdlc-outcome-validator",
  "version": "1.0.0",
  "status": "pass | pass_with_caveats | fail",
  "metrics_evaluated": 5,
  "metrics_passed": 5,
  "metrics_partial": 0,
  "metrics_failed": 0,
  "loop_back_count": 0,
  "caveats": [],
  "validation_timestamp": "<ISO-8601>"
}
```

## Loop-Back Rules

1. **Maximum 2 iterations** -- if a metric still fails after 2 loop-backs, mark the output as `fail` and deliver with explicit failure documentation
2. **Targeted re-execution** -- only re-execute the specific step that failed, not the entire workflow
3. **Progress tracking** -- each loop-back increments `loop_back_count` and records what was re-executed
4. **No infinite loops** -- the 2-iteration cap is mandatory and cannot be overridden

## Agent Integration

Agents reference this skill in their Skills section:
```
- sdlc-outcome-validator (success metrics validation loop)
```

The agent's existing `## Success Metrics` section defines the criteria. The agent's final workflow step invokes this skill's validation loop pattern.

## Validation Execution Example

```
=== Outcome Validation ===

METRIC: "Every finding references a specific file and line range"
CHECK:  S