Create structured build plans from feature requests or Flow IDs. Use when planning features or designing implementation. Triggers on /flow-next:plan with text descriptions or Flow IDs (fn-1, fn-1.2).
Turn a rough idea into an epic with tasks in .flow/. This skill does not write code.
Follow this skill and linked workflows exactly. Deviations cause drift, bad gates, retries, and user frustration.
IMPORTANT: This plugin uses .flow/ for ALL task tracking. Do NOT use markdown TODOs, plan files, TodoWrite, or other tracking methods. All task state must be read and written via flowctl.
CRITICAL: flowctl is BUNDLED — NOT installed globally. which flowctl will fail (expected). Always use:
ROOT="$(git rev-parse --show-toplevel)"
PLUGIN_ROOT="$ROOT/plugins/flow-next"
FLOWCTL="$PLUGIN_ROOT/scripts/flowctl"
$FLOWCTL <command>
Role: product-minded planner with strong repo awareness.
Goal: produce an epic with tasks that match existing conventions and reuse points.
Task size: every task must fit one /flow-next:work iteration. If it won't, split it.
Full request: $ARGUMENTS
Accepts:
fn-N to refine existing epicfn-N.M to refine specific taskExamples:
/flow-next:plan Add OAuth login for users/flow-next:plan fn-1/flow-next:plan fn-1 then review via /flow-next:plan-reviewIf empty, ask: "What should I plan? Give me the feature or bug in 1-5 sentences."
Check available backends and configured preference:
HAVE_RP=0
if command -v rp-cli >/dev/null 2>&1; then
HAVE_RP=1
elif [[ -x /opt/homebrew/bin/rp-cli || -x /usr/local/bin/rp-cli ]]; then
HAVE_RP=1
fi
# Check configured backend (priority: env > config)
CONFIGURED_BACKEND="${FLOW_REVIEW_BACKEND:-}"
if [[ -z "$CONFIGURED_BACKEND" ]]; then
CONFIGURED_BACKEND="$($FLOWCTL config get review.backend 2>/dev/null | jq -r '.value // empty')"
fi
MUST RUN the detection command above and use its result. Do not assume rp-cli is missing without running it.
Parse the arguments for these patterns. If found, use them and skip questions:
Research approach (only if rp-cli available):
--research=rp or --research rp or "use rp" or "context-scout" or "use repoprompt" → context-scout--research=grep or --research grep or "use grep" or "repo-scout" or "fast" → repo-scoutReview mode:
--review=opencode or "opencode review" or "use opencode" → OpenCode review (GPT-5.2, reasoning high)--review=rp or "review with rp" or "rp chat" or "repoprompt review" → RepoPrompt chat (via flowctl rp chat-send)--review=export or "export review" or "external llm" → export for external LLM--review=none or --no-review or "no review" or "skip review" → no reviewIMPORTANT: Ask setup questions in plain text only. Do NOT use the question tool. This is required for voice dictation (e.g., "1a 2b").
Skip review question if: Ralph mode (FLOW_RALPH=1) OR backend already configured (CONFIGURED_BACKEND not empty). In these cases, only ask research question (if rp-cli available):
Quick setup: Use RepoPrompt for deeper context?
a) Yes, context-scout (slower, thorough)
b) No, repo-scout (faster)
(Reply: "a", "b", or just tell me)
If rp-cli not available, skip questions entirely and use defaults.
Otherwise, output questions based on available backends:
If rp-cli available:
Quick setup before planning:
1. **Research approach** — Use RepoPrompt for deeper context?
a) Yes, context-scout (slower, thorough)
b) No, repo-scout (faster)
2. **Review** — Run Carmack-level review after?
a) Yes, OpenCode review (GPT-5.2, reasoning high)
b) Yes, RepoPrompt chat (macOS, visual builder)
c) Yes, export for external LLM (ChatGPT, Claude web)
d) No
(Reply: "1a 2a", "1b 2d", or just tell me naturally)
If rp-cli not available:
Quick setup before planning:
**Review** — Run Carmack-level review after?
a) Yes, OpenCode review (GPT-5.2, reasoning high)
b) Yes, export for external LLM
c) No
(Reply: "a", "b", or just tell me naturally)
Wait for response. Parse naturally — user may reply terse ("1a 2b") or ramble via voice.
Defaults when empty/ambiguous:
grep (repo-scout)opencode, else rp if available, else noneIf rp-cli not available: skip research questions, use repo-scout, review defaults to opencode or none if disabled.
Defaults when no review backend available:
grepnoneRead steps.md and follow each step in order. The steps include running research subagents in parallel via the Task tool. If user chose review:
/flow-next:plan-review after Step 4, fix issues until it passes/flow-next:plan-review with export mode after Step 4All plans go into .flow/:
.flow/epics/fn-N.json + .flow/specs/fn-N.md.flow/tasks/fn-N.M.json + .flow/tasks/fn-N.M.mdNever write plan files outside .flow/. Never use TodoWrite for task tracking.
.flow/