/validate

Check a MECE decomposition for structural integrity, MECE compliance, and Agent SDK readiness.

Usage

/validate <decomposition JSON or file path>

Examples:

• /validate then paste decomposition JSON
• /validate output.json
• /validate after a /decompose or /interview session (validates the last output)

Validation Layers

1. Structural Validation (deterministic)

uv run mece-decomposer/skills/mece-decomposer/scripts/validate_mece.py <decomposition.json>

Checks:

• Schema compliance (required fields, types, valid enums)
• Hierarchical ID consistency (parent-child prefix pattern)
• Cross-branch dependency validity (IDs exist, no self-references)
• Fan-out limits (2-7 children per branch, max 7 parallel)
• Atom completeness (all atoms have atom_spec, all branches have orchestration)
• Prompt/tool limits (flag atoms with >5 tools or >500 word prompts)
• Depth limits (warn at >5 levels)

2. MECE Quality Assessment (judgment-based)

Using the scoring rubrics from the mece-decomposer skill:

• ME testing: definition-based, example-based, boundary-case at each level
• CE testing: scenario enumeration, negation test, stakeholder test
• Depth-adaptive rigor: L1 full, L2 pairwise, L3 spot-check, L4+ trust
• Weighted score aggregation

Output

## Validation Report

**Status:** PASS / FAIL
**ME Score:** 0.XX -- [interpretation]
**CE Score:** 0.XX -- [interpretation]
**Overall:** 0.XX -- [interpretation]

### Issues
| Severity | Location | Type | Description |
|----------|----------|------|-------------|
| error    | node:1.2 | gap  | Missing ... |
| warning  | node:2   | depth| Tree depth ... |

### Recommendations
- [Fix for each error]
- [Improvement for each warning]

Interactive Report (when MCP server connected)

The mece-validate MCP tool renders a visual report with clickable issue locations that navigate to the problem node in the tree.

Score Interpretation

Range	ME	CE
0.85-1.0	Strong: no overlap	Strong: no gaps
0.70-0.84	Acceptable: minor boundary issues	Acceptable: minor gaps documented
0.50-0.69	Weak: redefine boundaries	Weak: add missing components
< 0.50	Failed: re-cut this level	Failed: restructure

Quality gate: >= 0.70 for export, >= 0.85 for confidence.

Next Steps

• Score too low? "Let me fix the issues" -> re-run /validate
• Ready? "Export as Agent SDK code" -> /export