en un clic
meow-plan
// 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.
// 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.
Heavy-ceremony orchestration for big work — research, planning, adversarial review, phased implementation, audit, delivery. Use when the user says "epic", "long task", "build this end to end", or wants a feature that runs all night.
Use when the user starts an interactive-mode request with `/meow-archive`; archive or delete the current MeowFlow thread and proposal.
Use when the user starts an interactive-mode request with `/meow-code`; implement the current MeowFlow plan and append a code handoff.
Use when the user starts an interactive-mode request with `/meow-execute`; execute script, data, automation, or artifact work and append an execution handoff.
Use when the user starts an interactive-mode request with `/meow-flow` or `/mfl`; coordinate staged MeowFlow agents through mfl thread state, thread updates, mfl-run stage launches, and handoffs.
Use when the user starts an interactive-mode request with `/meow-review`; review current MeowFlow changes and append a review handoff.
| 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.
/meow-plan ... is not permission to plan locally before MeowFlow state is
checked. Always go through mfl first.mfl run --stage plan ... and stop after reporting the launched
agent details.mfl agent update-self
confirms this session is already the plan stage agent for the current thread.Strip the /meow-plan prefix. Treat remaining text as the planning request.
If running inside a Paseo agent, run mfl agent update-self; otherwise this
chat is not a confirmed plan stage agent.
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.
If status is repository-root, tell the user to create a linked worktree
with mfl worktree new and stop.
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>"
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.
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.
Only follow this section after mfl agent update-self confirms this chat is
the current plan stage agent.
Run mfl status.
Read mfl thread status <id> --no-color for the current thread.
Read recent handoffs with mfl handoff get -n 5 when they exist.
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]+)*$: lowercase ASCII letters and digits separated by
single hyphens, with no underscores, spaces, leading hyphen, or trailing
hyphen.
Persist the name with the CLI before creating proposal artifacts:
mfl thread set name '<name>'
If a concurrent mfl update causes a transient SQLite write conflict, wait
for the previous mfl command to finish and retry the thread-name write
once before proceeding.
After setting or changing the thread name, rename the current Paseo agent to
the named-thread format when PASEO_AGENT_ID is available:
paseo agent update "$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.
If the repository uses OpenSpec, create the matching OpenSpec change with the same name. Prefer project-local OpenSpec proposal skills or commands.
Keep planning scoped. Do not edit production code.
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.
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.