# Validate Schema

> Verifies data structures, files, YAML blocks, or agent output against their defined schema, producing a pass/fail report with actionable fix instructions.

- Skill: `auto-skiller/validate-schema` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add auto-skiller/validate-schema`
- Raw SKILL.md: https://api.skillmd.com/api/skills/auto-skiller/validate-schema/raw
- Safety review: PASS (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools, Testing & QA
- Tags: Data Validation, Schema Validation, Structural Contract, Yaml
- Author: Auto-Skiller (https://skillmd.com/u/auto-skiller)
- Updated: 2026-08-22
- Page: https://skillmd.com/skills/auto-skiller/validate-schema

---


# Validate Schema

This skill allows the agent to verify that any data structure, file, YAML block, or agent output conforms to its defined schema or structural contract — producing a precise, actionable pass/fail report before the data is acted upon or written to disk.

## Steps

1. **Locate and read the schema.** Find the authoritative schema definition:
   - For workspace YAML files → check `.db/.schemas/` for the appropriate schema
   - For toolbox files → check the toolbox `yaml_path` file's structure
   - For custom structures → derive from the written spec or identity file
2. **Read the target fresh.** Apply Zero-Drift Audit Law — read the actual current file/data live, not from cached context.
3. **Map required fields.** List every required field from the schema. Note which are optional.
4. **Check field presence.** For each required field, verify it exists in the target.
5. **Check field types.** Verify data types match (string vs int, list vs scalar, etc.).
6. **Check value constraints.** Validate enum values, non-empty lists, valid paths, valid statuses, etc.
7. **Check structural contract.** Verify nesting depth, key naming conventions, and ordering where prescribed.
8. **Compile violations.** For every failure, record: field path, expected value/type, actual value/type, severity (blocking vs warning).
9. **Determine verdict.**
   - Zero violations → **VALID ✅**
   - Warnings only → **VALID WITH WARNINGS ⚠️**
   - Any blocking violation → **INVALID ❌**
10. **Generate fix instructions.** For each violation, produce a specific, executable correction.
11. **Output** the validation report.

## Output Format

```yaml
validation_report:
  target: "<file or data description>"
  schema_source: "<where the schema was found>"
  verdict: "VALID | VALID_WITH_WARNINGS | INVALID"
  violations:
    - field_path: "<e.g. content.skills[0].maturity>"
      expected: "<expected type or value>"
      actual: "<what was found>"
      severity: "blocking | warning"
      fix: "<exact correction to apply>"
  summary: "<one-line verdict summary>"
```

## Notes
- **Always validate before writing.** Any automated file write should be preceded by a schema validation of the content.
- If no schema exists for a target, flag this as a gap and generate a schema proposal before proceeding.
- Validation failures on `.db/.system.board.yaml` or router files are critical — they corrupt agent routing.
- Pair with `planning` toolbox: if INVALID, generate a fix plan before retrying.