name: validate-all description: Run all X-DEEP validators (state, templates, skills), plus lint and typecheck across projects, report findings user_invocable: true triggers:

• validate
• run validators
• check the archi
• valide tout

/validate-all — Run all X-DEEP validators

Single entry point to verify the X-DEEP architecture is healthy. Owned by the xdeep-engineering agent.

When this skill is invoked

• Before any commit touching .agent/ or .claude/skills/
• Before opening a PR with auto-fix, enhancement, or blueprint label
• On demand when the user suspects drift
• By /scaffold as part of its step 4

What it runs

Core validators (always)

node .agent/scripts/validate-all.mjs

This runs:

• validate-state.mjs — checks .agent/state.json structure, no duplicates
• validate-templates.mjs — all .agent/templates/*.yaml parse + schema + cross-ref state
• validate-skills.mjs — all .claude/skills/*/SKILL.md frontmatter valid + unique names

Project checks (if relevant files changed)

• telegram-bot/ changed → cd telegram-bot && npm run lint
• dashboard/ changed → cd dashboard && npm run lint && npx tsc --noEmit && npm test
• sales-tool/ changed → cd sales-tool && npm run lint

Expected output format

Success case:

✅ All validators passed

Core:
- state.json: OK
- templates: OK (10 agents)
- skills: OK (15 skills)

Projects checked: telegram-bot (lint OK)

Failure case:

❌ Validation failed (2 errors)

Core errors:
- .agent/templates/xdeep-sales.yaml: reports_to references non-existent agent
- .claude/skills/foo/SKILL.md: frontmatter missing required field 'name'

Project errors:
- telegram-bot lint: 3 warnings, 1 error in src/index.mjs

Suggested next step: fix errors above, then re-run /validate-all

Guardrails

• NEVER commit if validation fails
• If validator script itself errors (not a validation failure): escalate to the user
• If a fix requires changing schemas: stop and propose, don't modify schemas autonomously