Process: Onboarding a New Factory Worker Role

Process: Onboarding a New Factory Worker Role

Owner: Training Specialist / HR (Tara) Type: Manual Frequency: Per new role deployment (triggered by Role Factory completion)

Purpose

Defines the complete onboarding process for a newly deployed factory worker role. Ensures that every role is verified, validated, and confirmed operational before it is marked as Active. No role goes live without completing this process.

Trigger

Role Factory completes Stage 6 (DEPLOY) for a new role — all role files exist and the test report score meets the deployment threshold.

Steps

1. Verify role files

Actions:

  • Confirm all required files exist for the role type:
    • role.md — present and follows the standard template
    • eval-cases.md — present with 20 triggers, 20 non-triggers, 3 scenarios
    • test-report.md — present with score >= 13/15
    • For Agent-type roles: agent.md exists, slash commands exist in .claude/commands/
    • For Process-type roles: processes/ directory exists with documented processes
  • Verify naming conventions follow Nomenclature Specialist (Nora) standards
  • Verify directory placement matches org chart department

Output: File verification checklist (pass/fail per item).

Gate: All required files must be present. If any are missing, return to Role Factory for remediation.

2. Validate role definition

Actions:

  • Read role.md thoroughly
  • Verify responsibility statement is clear and unambiguous
  • Verify inputs table lists all expected sources with formats
  • Verify outputs table lists all deliverables with destinations
  • Verify interactions table covers all roles this role touches
  • Verify success criteria are measurable and have targets
  • Cross-reference with org chart to confirm department, phase, and type match

Output: Role definition validation report (pass/fail per section with notes).

Gate: Role definition must be complete and internally consistent.

3. Validate commands or processes

Actions:

  • For Agent-type roles:
    • Run each slash command with a representative input
    • Verify command produces expected output format
    • Verify command references correct files and paths
  • For Process-type roles:
    • Read each process document
    • Verify each process has: purpose, trigger, steps, outputs, quality gate
    • Verify process steps are actionable (not vague)
    • Verify handoffs to other roles are clearly defined

Output: Command/process validation results.

Gate: All commands must execute without errors; all processes must be complete and actionable.

4. Run eval cases

Actions:

  • Present each of the 20 trigger phrases and verify:
    • The role recognizes it as within scope
    • The role maps it to the correct process or command
  • Present each of the 20 non-trigger phrases and verify:
    • The role correctly identifies it as out of scope
    • The role can name the correct target role
  • Walk through each of the 3 scenarios and verify:
    • The role follows the expected behavior steps
    • Outputs match the expected quality

Output: Eval case results (pass/fail per case, notes on any failures).

Gate: >= 90% of trigger phrases correctly handled. >= 90% of non-trigger phrases correctly routed. All 3 scenarios pass quality check.

5. Confirm interactions work

Actions:

  • For each role listed in the interactions table:
    • Verify the counterpart role exists and is active (or is a known future role)
    • Verify the handoff format is agreed upon by both sides
    • If the counterpart is active, simulate one handoff in each direction
  • Identify any interaction gaps (e.g., counterpart role not yet built)
  • Document any deferred interactions (counterpart is Future status)

Output: Interaction validation matrix (role x role, status of each handoff).

Gate: All interactions with Active roles must be confirmed working. Deferred interactions documented but do not block onboarding.

6. Mark as onboarded

Actions:

  • Compile the full onboarding report:
    • File verification results
    • Role definition validation results
    • Command/process validation results
    • Eval case results
    • Interaction validation results
    • Overall readiness assessment: Ready / Conditionally Ready / Not Ready
  • If Ready or Conditionally Ready:
    • Create an onboarding checklist record for the role
    • Update the role lifecycle record (see role-lifecycle.md) to status: Active
    • Notify CTO (Clara) that the role is onboarded and operational
  • If Not Ready:
    • Document specific failures
    • Return to Role Factory or relevant owner for remediation
    • Re-run onboarding after fixes

Output: Completed onboarding report and checklist record.

Onboarding checklist template

# Onboarding Checklist: [Role Name] ([Persona])

- **Date:** [Date]
- **Onboarded by:** Tara (Training Specialist)
- **Overall status:** [Ready | Conditionally Ready | Not Ready]

## File verification
- [ ] role.md present and valid
- [ ] eval-cases.md present (20 triggers, 20 non-triggers, 3 scenarios)
- [ ] test-report.md present (score >= 13/15)
- [ ] Agent files present (agent.md, commands) — if applicable
- [ ] Process files present (processes/) — if applicable

## Role definition
- [ ] Responsibility is clear and unambiguous
- [ ] Inputs are complete with sources and formats
- [ ] Outputs are complete with destinations
- [ ] Interactions cover all touchpoints
- [ ] Success criteria are measurable

## Commands / processes
- [ ] All commands execute successfully — if applicable
- [ ] All processes are complete and actionable — if applicable

## Eval cases
- [ ] Trigger phrases: [X]/20 pass (>= 18 required)
- [ ] Non-trigger phrases: [X]/20 pass (>= 18 required)
- [ ] Scenarios: [X]/3 pass (3/3 required)

## Interactions
- [ ] All Active-role interactions confirmed working
- [ ] Deferred interactions documented: [list or "none"]

## Notes
[Any conditions, caveats, or follow-up items]

Quality gate

An onboarding is complete when:

  • All 6 steps have been executed
  • Onboarding checklist is filled out and stored
  • Role lifecycle record is updated
  • CTO has been notified of the result