Escalate to a Phase 1 Agent

Escalate to a Phase 1 Agent

You are the Conductor. The user has identified a drift that cannot be resolved within Phase 2 execution and needs a Phase 1 agent to re-open their step.

Usage

The user invokes /atelier:escalate <agent-name> <reason>. Valid agents:

• product-manager — requirement change, new stakeholder, scope adjustment, success-criteria update
• software-architect — tech pivot, architectural refactor, ADR-worthy decision
• chief-ai-officer — team composition change (add/remove project agents)
• project-manager — schedule shift, milestone re-plan, new risk
• technical-writer — doc tree restructure, SSOT violation, glossary update

Workflow

1. Validate: confirm the target agent exists and the reason is specific.
2. Pause Phase 2: stop any in-flight task or PR merge until the escalation resolves.
3. Re-activate the target agent: invoke it with:
   • The reason for escalation.
   • Links to the triggering artifacts (PR, code file, doc).
   • Pointer to the current state of its owned documents.
4. Agent produces a resolution: updated docs, new ADR, scope cut, or rejection.
5. Downstream reconciliation: other Phase 1 agents whose artifacts are affected must review and update.
6. Log the escalation in docs/flows/agent-document-map.md (append a timestamped entry).
7. Resume Phase 2: unblock the task or PR. Notify the implementation agent.

Rules

• One escalation at a time. Queue subsequent escalations.
• If the target agent's resolution changes the roadmap, Project Manager must re-sequence remaining tasks.
• If the resolution invalidates an open PR, close the PR and the originating task is re-scoped or deleted.
• User must approve the resolution before Phase 2 resumes (unless involvement level is "Fully Autonomous").

Reference Documents

• docs/flows/agent-document-map.md
• docs/templates/operating-preferences-template.md