meow-plan

name: meow-plan description: Use when the user starts an interactive-mode request with /meow-plan; use mfl to update/discover the thread, launch a plan stage agent when needed, and create the matching proposal without editing production code.

Launch or run the planner role for a MeowFlow thread.

Entry Contract

• /meow-plan ... is not permission to plan locally before MeowFlow state is checked. Always go through mfl first.
• When invoked from the middle of an existing non-MeowFlow agent chat, spawn a planner with mfl run --stage plan ... and stop after reporting the launched agent details.
• Only perform planning work in the current chat when mfl agent update-self confirms this session is already the plan stage agent for the current thread.

Coordinator Flow

1. Strip the /meow-plan prefix. Treat remaining text as the planning request.

2. If running inside a Paseo agent, run mfl agent update-self; otherwise this chat is not a confirmed plan stage agent.

3. If mfl agent update-self confirms this session is the plan stage agent, skip to Plan Stage Flow. Otherwise run mfl status and stay in coordinator-mode.

4. If status is repository-root, tell the user to create a linked worktree with mfl worktree new and stop.

5. If the user provided a planning request, launch the planner through mfl instead of doing the work locally. Preserve --provider <provider> when the entry command included it:

   mfl run --stage plan "<planning-request>"
   mfl run --stage plan --provider <provider> "<planning-request>"
   

6. Report the returned thread-id, worktree, provider, agent-id, and next-seq when mfl run prints them, then direct the user to continue in the new agent chat.

7. If no planning request was provided and the current chat is not the plan stage agent, report the current thread from mfl status and ask for the plan request. Do not create proposal files locally.

Plan Stage Flow

Only follow this section after mfl agent update-self confirms this chat is the current plan stage agent.

1. Run mfl status.

2. Read mfl thread status <id> --no-color for the current thread.

3. Read recent handoffs with mfl handoff get -n 5 when they exist.

4. Choose a readable unused kebab-case thread name such as install-meow-flow-skills, unless the user explicitly requested an existing branch or name. Thread names must match ^[a-z0-9]+(-[a-z0-9]+)*$: lowercase ASCII letters and digits separated by single hyphens, with no underscores, spaces, leading hyphen, or trailing hyphen.

5. Persist the name with the CLI before creating proposal artifacts:

   mfl thread set name '<name>'
   

   If a concurrent mfl update causes a transient SQLite write conflict, wait for the previous mfl command to finish and retry the thread-name write once before proceeding.

6. After setting or changing the thread name, rename the current Paseo agent to the named-thread format when PASEO_AGENT_ID is available:

   paseo agent update "$PASEO_AGENT_ID" --name "Plan: <name> (1)"
   

   Use the current stage label and the agent's sequence from mfl thread status <id> --no-color when the renamed agent is not the first plan agent.

7. If the repository uses OpenSpec, create the matching OpenSpec change with the same name. Prefer project-local OpenSpec proposal skills or commands.

8. Keep planning scoped. Do not edit production code.

9. Commit proposal artifacts. Respect git config --local paseo.prompt.title when set; otherwise infer the local title style from recent main-branch commits. Use docs: add proposal <name> as the fallback.

10. Append a concise planning handoff with the next suggested stage, usually /mfl code or /mfl execute:

mfl handoff append --stage plan "planned <name>; proposal <path>; next <stage>"

The coding and review stages stay idle until the user approves or asks for the next stage.