work-item-review-output-formatlisted

# Work-Item Review Output Format ## JSON Schema Return your analysis as a JSON code block. Do not include any text before or after the JSON block — the orchestrator will parse this output directly. ```json { "lens": "<lens-identifier>", "summary": "2-3 sentence assessment from this lens perspective.", "strengths": [ "Positive observation about what the work item gets right from this lens perspective" ], "findings": [ { "severity": "critical", "confidence": "high", "lens": "<lens-identifier>", "location": "Acceptance Criteria", "title": "Brief finding title", "body": "🔴 **<Lens Name>**\n\n[Issue description — 1-2 sentences with enough context to understand standalone].\n\n**Impact**: [Why this matters — 1 sentence].\n\n**Suggestion**: [Concrete fix — 1-2 sentences]." } ] } ``` An optional `synthetic` boolean field may appear on findings. It is set to `true` only by the orchestrator's malformed-agent fallback — agents must never emit it. ## Field Reference - **lens**: Agent lens identifier (e.g., `"completeness"`, `"scope"`, …) - **summary**: 2-3 sentence assessment from this lens perspective. Reflect the key dimensions from the lens's Core Responsibilities. This is where holistic assessment lives, beyond individual findings. - **strengths**: Positive observations (fed into the review summary — never posted as individual findings) - **findings**: All findings, each referencing a location in the work item