pm%3aplan（下午%3a 计划）

Compatibility: AskUserQuestion falls back to numbered list on non-Claude-Code platforms.

Skill Requirements

command -v npx >/dev/null 2>&1 || echo "MISSING: npx"

Critical Rules

• Never generate PRDs — pm:plan writes roadmaps only; use pm:prd for PRDs
• Do not make engineering decisions, focus on product development
• Always write to .codevoyant/roadmaps/ first (draft); use pm:approve to commit
• No markdown tables — use bullets, definition lists, or Mermaid diagrams
• Capability tiers group features by strategic value, not by team or sprint
• Research context from pm:explore should be consumed if available
• Every capability's "Why now" must cite a source — research artifact, competitive signal, or user data. If none is available, mark it explicitly as [ASSUMPTION — unvalidated]
• Do not generate a roadmap from strategic intent alone — always check .codevoyant/explore/ for available research context before drafting
• Capabilities without any research backing must be placed in Tier 3 (Future) and flagged

Step 0: Parse arguments

HORIZON="${1:-}"   # quarter | half | annual
BG_FLAG=false; SILENT=false
[[ "$*" =~ --bg|-b ]] && BG_FLAG=true
[[ "$*" =~ --silent ]] && SILENT=true

Step 1: Gather context

Before asking, scan for available research:

ls .codevoyant/explore/ 2>/dev/null | head -20

Store discovered directory names as EXPLORE_DIRS. If none found, set EXPLORE_DIRS="(none)".

Ask:

AskUserQuestion:
  questions:
    - question: "What time horizon is this roadmap for?"
      header: "Horizon"
      options:
        - label: "Quarter (3 months)"
        - label: "Half-year (6 months)"
        - label: "Annual (12 months)"
    - question: "Is there research context to pull in?"
      header: "Research"
      options:
        - label: "Yes — use .codevoyant/explore/ (found: {EXPLORE_DIRS})"
        - label: "No — start from scratch"
    - question: "What is the primary strategic goal for this period?"
      header: "Goal"
      options:
        - label: "I'll describe below"
        - label: "Pull from existing product docs"

If using research: list dirs in .codevoyant/explore/ and ask which to include. Read summary.md and all files under research/ in selected dirs as RESEARCH_CONTEXT.

Step 1.5: Research backfill (if no research artifacts)

After the user selects "Yes — use.codevoyant/explore/", check if that directory actually has content.

If research artifacts exist: load selected ones as RESEARCH_CONTEXT and skip this step.

If no research exists (or user selected "No — start from scratch"):

Tell the user: "No research context found — I'll run a quick web search to ground the roadmap."

Ask one round of clarifying questions:

AskUserQuestion:
  questions:
    - question: "What product category or market is this roadmap for?"
      header: "Market"
      options:
        - label: "I'll describe below"
    - question: "Who are your top 2–3 known competitors (or 'unknown')?"
      header: "Competitors"
      options:
        - label: "I'll list below"
        - label: "Unknown — research it"
    - question: "What is the biggest user problem you're solving this period?"
      header: "Core problem"
      options:
        - label: "I'll describe below"

Launch 3 Sonnet agents in parallel (run_in_background: false, model: claude-sonnet-4-6):

Agent A — Market signals: Prompt: Search for market trends and user needs in "{PRODUCT_CATEGORY}". Run WebSearch("{PRODUCT_CATEGORY} market trends {year}"), WebSearch("{PRODUCT_CATEGORY} user needs"), WebSearch("{PRODUCT_CATEGORY} industry report"). Fetch 2 relevant URLs. Write 5 key findings to .codevoyant/explore/{SLUG}/research/market.md with citations and Tier labels.

Agent B — Competitive landscape: Prompt: Research competitors for "{PRODUCT_CATEGORY}". Run WebSearch("{PRODUCT_CATEGORY} competitors"), WebSearch("{COMPETITOR_LIST} product updates {year}"), WebSearch("{PRODUCT_CATEGORY} product launches"). Fetch 2 relevant URLs. Write findings to .codevoyant/explore/{SLUG}/research/competitive.md. Include each competitor's recent moves and strategic direction.

Agent C — Internal prior art: Prompt: Scan this repository for product context: read all files in docs/product/ and docs/prd/, read any.codevoyant/roadmaps/ files, note existing strategic goals, past PRDs, known user problems. Write a summary to .codevoyant/explore/{SLUG}/research/internal.md.

Wait for all three to complete. Read outputs as RESEARCH_CONTEXT.

Executive decision protocol for gaps: When research is absent or ambiguous, make confident strategic calls rather than hedging. The guiding question: *what is most likely to create a sticky, resonant product that users will keep coming back to?* Principles:

• Prioritize capabilities that create strong user habits or network effects over one-time utility
• Prefer fewer, deeper bets over a long list of incremental features
• Choose differentiation over parity — what makes this product worth choosing over alternatives?
• Place bets on user outcomes, not feature completeness

Label every such decision [DESIGN DECISION] in the roadmap. Do not leave "Why now" blank or write "strategic priority" without substance — make a call and explain it briefly.

Step 2: Confirm scope

Present a one-paragraph summary (horizon, goal, research sources), then immediately proceed to Step 3. Do not ask a confirmation question here — the user will review and adjust the draft in Step 4.

Step 3: Draft roadmap

Set:

• DATE = current date YYMMDD
• TYPE = quarter | half | annual (from HORIZON)
• DRAFT_PATH = .codevoyant/roadmaps/{DATE}-{TYPE}-roadmap.md

Use the Agent tool to spawn the roadmap drafter:

• pm-planner (see agents/pm-planner.md) — drafts the roadmap using research context and strategic goal

Wait for completion. Read DRAFT_PATH.

Step 4: Review draft

Present the draft to the user. Ask:

AskUserQuestion:
  question: "Does this roadmap draft look right?"
  header: "Draft review"
  options:
    - label: "Accept — run pm:review"
    - label: "Adjust (describe changes)"
    - label: "Discard"

Loop until accepted or discarded. If adjustments needed, apply them to DRAFT_PATH and re-present.

Step 5: Run pm:review

After acceptance, run /pm:review on the draft to catch quality issues automatically.

Step 6: Register + Notify

Register the roadmap draft:

npx @codevoyant/agent-kit plans register \
  --name "{DATE}-{TYPE}" \
  --plugin pm \
  --description "{strategic goal first line}" \
  --total "0"

(--total 0 — pm roadmaps have no implementation tasks; those are created by em:plan)

if [ "$SILENT" != "true" ]; then
  npx @codevoyant/agent-kit notify \
    --title "pm:plan complete" \
    --message "Draft roadmap written to {DRAFT_PATH}. Use /pm:approve to commit."
fi

Report: "Draft written to {DRAFT_PATH}. Run /pm:approve when ready to commit to docs/product/."