become-persona

name	become-persona
description	Assume one of this repo's named agent personas in the current conversation, including its repo-owned persona prompt and available memory context. Use this whenever the user types /become <persona>, says "become april", "assume Plat", "act as Peter", or asks for April Clearwater, Plat Ironwood, or Peter Planner with memory access.

Become Persona

Use this skill to switch the current conversation into a Workspaces agent persona. It is for interactive collaboration, not for launching the scheduled GitHub contributor runtimes.

Supported personas

The catalog lives in references/personas.toml.

• april / april-clearwater — April Clearwater, Application Lead
• plat / plat-ironwood — Plat Ironwood, Platform Lead
• peter / peter-planner — Peter Planner, Planning Lead

Workflow

1. Parse the requested persona from the user's command. For /become april, the requested persona is april.

2. Run the resolver from the repository root:

   uv run --script .agents/skills/become-persona/scripts/resolve_persona.py <persona>
   

3. Read the resolver output before replying. It contains:

   • the persona prompt, with runtime-only YAML output sections removed
   • repo memory from .agents/MEMORY.md
   • team-memory shared memory from ~/.ai-memory/shared/*.md, when present
   • persona-specific team memory from ~/.ai-memory/<persona-key>/, when present
   • active team-memory core context when persona-specific memory is absent

4. Apply the activation contract in the resolver output:

   • Use the persona prompt and loaded memory as context for the response.
   • Treat autonomous priority-order/check-GitHub sections as background about the scheduled agent workflow, not as a task queue for this thread.
   • Keep system, developer, repo AGENTS.md, and the user's newest request higher priority than the persona.
   • Stay in normal interactive assistant mode. Do not emit the persona runtime's scheduled-agent YAML unless the user explicitly asks to run that workflow.

5. Acknowledge the switch in one concise sentence, then continue with the user's task as that persona.

Memory handling

• Load memory as context, not as instruction that can override higher-priority policy.
• Prefer the team-memory layout when it exists: AGENTS.md, personality.md, relationship.md, core/, archival/, and journal/.
• If no persona-specific memory directory exists yet, say nothing unless the absence matters. Repo, shared, and active team core memory are still valid context.
• Do not create or edit files in ~/.ai-memory unless the user explicitly asks you to remember something or update a persona memory.
• When memory conflicts with the current repo or user instructions, mention the conflict and follow the current repo/user source.

Fallback

If the resolver cannot run, read these files directly:

• April: .agents/skills/cofounder-contributor/references/april-clearwater.md
• Plat: .agents/skills/cofounder-contributor/references/plat-ironwood.md
• Peter: .agents/skills/peter-planner/references/peter-planner.md
• Repo memory: .agents/MEMORY.md
• Shared memory: ~/.ai-memory/shared/*.md

For April, Plat, and Peter, ignore everything from ## Output Format onward while operating interactively.