flow-next-work

Flow work

Execute a plan systematically. Focus on finishing.

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.

Preamble

CRITICAL: flowctl is BUNDLED — NOT installed globally. which flowctl will fail (expected). Define once; subsequent blocks (here and in phases.md) use $FLOWCTL:

FLOWCTL="${DROID_PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT}}/scripts/flowctl"
[ -x "$FLOWCTL" ] || FLOWCTL=".flow/bin/flowctl"

Pre-check: Local setup version

Non-blocking, same pattern as /flow-next:plan — one-line nag when the local setup lags the plugin:

SETUP_VER=$(jq -r '.setup_version // empty' .flow/meta.json 2>/dev/null)
PLUGIN_JSON="${DROID_PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT}}/.claude-plugin/plugin.json"
PLUGIN_VER=$(jq -r '.version' "$PLUGIN_JSON" 2>/dev/null || echo "unknown")
if [[ -n "$SETUP_VER" && "$PLUGIN_VER" != "unknown" && "$SETUP_VER" != "$PLUGIN_VER" ]]; then
  echo "Plugin updated to v${PLUGIN_VER}. Run /flow-next:setup to refresh local scripts (current: v${SETUP_VER})." >&2
fi

Continue regardless (never blocks; silent when setup was never run or versions match).

Hard requirements (non-negotiable):

• You MUST run flowctl done for each completed task and verify the task status is done.
• You MUST stage with git add -A (never list files). This ensures .flow/ and scripts/ralph/ (if present) are included.
• Do NOT claim completion until flowctl show <task> reports status: done.
• Do NOT invoke /flow-next:impl-review until tests/Quick commands are green.

Role: execution lead, plan fidelity first. Goal: complete every task in order with tests.

Ralph Mode Rules (always follow)

If REVIEW_RECEIPT_PATH is set or FLOW_RALPH=1:

• Must use flowctl done and verify task status is done before committing.
• Must stage with git add -A (never list files).
• Do NOT use TodoWrite for tracking.

Autonomous Mode (questions off, no receipt obligations)

If $ARGUMENTS contains the literal token mode:autonomous (strip it — same parse shape as capture's mode:autofix, a NEW branch) or FLOW_AUTONOMOUS=1 is set:

• Ask NO setup questions (branch + review questions below are suppressed).
• Branch defaults deterministically to --branch=new when no explicit branch option is present — under autonomy "the user's answer" never exists, and defaulting to the current branch could commit straight to main. Name the new branch exactly the spec's branch_name field ($FLOWCTL show <spec-id> --json | jq -r '.branch_name') — pilot's branch matrix, its all-done PR probe, and make-pr's branch-match spec detection all key on that name; an ad-hoc name breaks multi-tick continuity.
• Review = explicit --review passthrough if present, else the configured backend (none when REVIEW_BACKEND is ASK).
• Autonomy ≠ Ralph. Neither signal sets FLOW_RALPH, implies REVIEW_RECEIPT_PATH receipt obligations, or activates ralph-guard hooks. The Ralph rules above apply only under their own markers (the done/git add -A/no-TodoWrite discipline is universal anyway).
• Never hang on a question. A genuinely unanswerable ambiguity → stop cleanly with a one-line NEEDS_HUMAN: <reason> report instead of asking.

Input

Full request: $ARGUMENTS

Accepts:

• Flow spec ID fn-N-slug (e.g., fn-1-add-oauth) or legacy fn-N/fn-N-xxx to work through all tasks
• Flow task ID fn-N-slug.M (e.g., fn-1-add-oauth.2) or legacy fn-N.M/fn-N-xxx.M to work on single task
• Markdown spec file path (creates spec from file, then executes)
• Idea text (creates minimal spec + single task, then executes)
• Chained instructions like "then review with /flow-next:impl-review"

Examples:

• /flow-next:work fn-1-add-oauth
• /flow-next:work fn-1-add-oauth.3
• /flow-next:work fn-1 (legacy formats fn-1, fn-1-xxx still supported)
• /flow-next:work docs/my-feature-spec.md
• /flow-next:work Add rate limiting
• /flow-next:work fn-1-add-oauth then review via /flow-next:impl-review

If no input provided, ask for it.

FIRST: Parse Options or Ask Questions

Check configured backend:

REVIEW_BACKEND=$($FLOWCTL review-backend)

Returns: ASK (not configured), or rp/codex/none (configured).

Option Parsing (skip questions if found in arguments)

Parse the arguments for these patterns. If found, use them and skip corresponding questions:

Branch mode:

• --branch=current or --current or "current branch" or "stay on this branch" → current branch
• --branch=new or --new-branch or "new branch" or "create branch" → new branch
• --branch=worktree or --worktree or "isolated worktree" or "worktree" → isolated worktree

Review mode:

• --review=codex or "review with codex" or "codex review" or "use codex" → Codex CLI (GPT 5.5 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 review

Autonomous mode:

• mode:autonomous token (stripped from arguments) or FLOW_AUTONOMOUS=1 env → suppress ALL setup questions; defaults per the Autonomous Mode section above (branch new, review = configured backend).

If options NOT found in arguments

If AUTONOMOUS=1 (autonomous mode): ask nothing — apply the autonomous defaults and continue to the workflow.

If REVIEW_BACKEND is rp, codex, or none (already configured): Only ask branch question. Show override hint:

Quick setup: Where to work?
a) Current branch  b) New branch  c) Isolated worktree

(Reply: "a", "current", or just tell me)
(Tip: --review=rp|codex|export|none overrides configured backend)

If REVIEW_BACKEND is ASK (not configured): Ask both branch AND review questions:

Quick setup before starting:

1. **Branch** — Where to work?
   a) Current branch
   b) New branch
   c) Isolated worktree

2. **Review** — Run Carmack-level review after?
   a) Codex CLI
   b) RepoPrompt
   c) Export for external LLM
   d) None (configure later with --review flag)

(Reply: "1a 2a", "current branch, codex", or just tell me naturally)

Wait for response. Parse naturally — user may reply terse or ramble via voice.

Defaults when empty/ambiguous:

• Branch = new
• Review = configured backend if set, else none (no auto-detect fallback)

Do NOT read files or write code until user responds.

Workflow

After setup questions answered, read [phases.md](phases.md) and execute each phase in order.

Worker subagent model: Each task is implemented by a worker subagent with fresh context. This prevents context bleed between tasks and keeps re-anchor info with the implementation. The main conversation handles task selection and looping; worker handles implementation, commits, and reviews.

If user chose review, pass the review mode to the worker. The worker invokes /flow-next:impl-review after implementation and loops until SHIP.

Completion review gate: When all tasks in a spec are done, if --require-completion-review is configured (via flowctl next), the work skill invokes /flow-next:spec-completion-review before allowing the spec to close. This verifies the combined implementation satisfies the spec. The spec-completion-review skill handles the fix loop internally until SHIP.

Tracker sync (opt-in, off by default)

The no-tracker path is the documented default and is behaviorally unchanged. Every tracker touchpoint below runs ONLY when the bridge is active AND the specific event is opted in; otherwise it is a silent no-op (no new steps, no new prerequisites). The bridge is act