create-agentflow-plan

name	create-agentflow-plan
description	Create an implementation plan from an approved AgentFlow architecture without adding or dropping scope.

Create AgentFlow Plan

Use this repo-local skill only after the relevant AgentFlow architecture is approved.

Contract

The plan must match the approved architecture exactly. Do not add behavior that is not in the approved architecture. Do not omit approved models, flows, document sections, decisions, jobs, events, permissions, or terminal outcomes.

Workflow

1. Read AGENTS.md, .agents/rules/, relevant .agents/memory/, and existing repository planning conventions.
2. Read the approved architecture:

python3 .agentflow/scripts/agentflow.py snapshot

3. Create the plan at the repository's existing plan location if one exists. Otherwise use .agentflow/plans/<proposal-or-feature-id>.md.
4. Structure the plan around the approved architecture:
   • app and model changes;
   • app-owned enum, mixin, admin, permission, signal, and other approved document changes;
   • flow-by-flow implementation steps;
   • API, function, job, event, integration, and permission work;
   • data migration or backfill work when required;
   • verification steps that prove the approved architecture was implemented.
5. Add a traceability checklist mapping every approved model entity, flow node or edge, and document section or entry to a plan item.
6. Audit the finished plan against the approved architecture. Update the plan until every approved architecture element is represented exactly once and no unapproved behavior remains.

Memory

If planning reveals a stable repository convention or reusable user preference, record it in .agents/memory/project-context.md. Do not store temporary task notes as memory.

Stop Point

After the plan is complete, report the plan path and the architecture revision or proposal it maps to. Do not implement unless the user has asked for implementation.