使い方
- インストールコマンドをコピーし、自分の環境へ導入する前にソースを確認します。
- 必要なツール、認証情報、対応環境を原文で確認します。
- 入力と成果物を指定し、出力を資料と照合します。
npx skills add https://github.com/uxcel-lab/product-skills --skill pm-roadmap開始前の確認
- 導入にはモデル契約、API枠、ファイルへのアクセス許可は含まれません。
- スター数や導入数は活動の指標であり、独立した品質・安全性試験ではありません。
リポジトリの原文
見出し、例、コード、表、リンク、参照画像を含む原文を表示しています。
Product Roadmap Skill
How this skill behaves (read first)
This is a generative skill. The default failure mode is the dated feature list: features pinned to months, presented as commitments, mistaken for a backlog, stale within a quarter — manufacturing an illusion of certainty that converts every market shift into a broken promise. A roadmap is a strategic communication tool: it answers why and what-outcome, in sequence, at the right confidence level. This skill gates:
- Establish strategy, audience, and horizon — these decide format, detail, and date precision.
- Apply the always-true core — outcomes over outputs, themes over features, strategy linkage, honest uncertainty.
- Surface the format and process decisions (Now-Next-Later vs. timeline, quarterly vs. annual, audience views) with trade-offs.
Then it hands off to `pm-prioritization-rigor-audit` and `pm-okr-metric-validity-audit` for validation.
Step 0 — Establish context before drafting
Ask if not known; state the assumption if proceeding without an answer:
- Is there a strategy and are there OKRs? A roadmap translates strategy into sequenced initiatives. No strategy → stop; help articulate goals first (a roadmap built without one is a wishlist). Existing key results become the spine: work backward from each KR to the initiatives believed to move it.
- Audience — internal teams (detail, dependencies, risks), executives (themes, milestones, strategic linkage), or customers (value and direction, no dates, no internals)? Each gets a different view of the same truth — never one document for all.
- Horizon and cadence — annual (budgeting, directional bets) vs. quarterly (execution detail)? What triggers replanning besides the calendar?
- Delivery context — fixed-date commitments (compliance, contracts) legitimately need timeline treatment; discovery-driven products need confidence-based structure.
The always-apply core (correct for almost every case)
Outcomes, not outputs
- Each roadmap item names the problem/outcome, not just the deliverable. Outputs (features shipped) measure activity; outcomes (behavior or business change) measure value.
- ❌ "Q2: AI assistant, dark mode, new dashboard" ✅ "Now: cut onboarding drop-off at step 3 (drives activation KR) — candidate bets listed under it"
- Translate business outcomes into product outcomes: "reduce churn" is too broad to act on; "increase week-1 feature adoption among new accounts" is a product outcome a team can own.
Themes and initiatives, not feature catalogues
- Theme = problem space tied to strategy ("onboarding," "localization"); initiative = significant effort within a theme; features live underneath as current bets. This keeps direction stable while solutions stay swappable.
- Every theme visibly traces to a strategic goal — an initiative that can't name its strategy link is a pet project in costume (the audit will flag it).
Honest uncertainty
- Confidence is part of the content. Near-term work is committed; far-out items are bets. Mark the difference explicitly — stakeholders treat everything undifferentiated as a promise.
- No false-precision dates. Internally, sequencing and dependency timing; externally, direction without delivery dates.
A living document with a communication job
- The roadmap is a shared tool, not a file — and it decays: schedule reviews (quarterly planning + event triggers: major discovery findings, competitor moves, strategy shifts) and feed continuous discovery into priorities (weekly user insights; dual-track discovery/delivery).
- Roadmap ≠ backlog. Strategic direction lives in the roadmap; tasks, bugs, and tickets live in the backlog. If items have story points, they're in the wrong document.
- Presentation is part of the deliverable: short, focused, visual (themes/colors/milestones), with a narrative thread — present challenges → outcomes → future vision. Answer "why is X delayed/excluded" with strategy, not apology.
The context-dependent decisions (surface, don't auto-apply)
| Decision | Apply when | Avoid / adapt when | Default recommendation |
|---|---|---|---|
| Now-Next-Later | Uncertain timing, sequence matters; discovery-driven teams | Contractual/compliance dates exist | The default modern format — confidence-organized |
| Theme-based view | Communicating direction to executives/customers | Teams needing execution sequencing beneath it | Themes on top, initiatives within, features as bets |
| Timeline/Gantt view | Hard dependencies, cross-team handoffs, fixed dates | As the public face (reads as a promise) | Internal execution layer only |
| Annual vs. quarterly | Annual: budget, hiring, directional bets. Quarterly: execution detail | One artifact for both jobs | Both, linked: annual themes → quarterly initiatives |
| Audience views | Internal-detailed / executive-thematic / customer-value | Showing customers internals, or execs ticket-level detail | Three views of one source of truth |
| Data-driven sizing | Opportunity sizing, impact estimates, historical timelines, success thresholds where analytics exist | Inventing numbers where data doesn't exist (→ rigor audit) | Size with real data; mark estimates as estimates |
| Dependency mapping | Multi-team initiatives, shared systems, external partners | Solo-team simple scopes | Map for anything crossing team boundaries |
| Technical-debt allocation | A standing capacity share for debt/platform work | Letting debt compete feature-by-feature (it always loses) | Explicit recurring allocation, themed like any other work |
| Stakeholder input process | Structured intake from sales/support/customers with bias awareness | Loudest-voice routing into the roadmap | Input channels in, strategy filter out; involve stakeholders at goal-setting to build buy-in |
Validate the result (orchestration)
Hand-offs name each lens by its installable skill name. Invoke one only if that skill is installed; if it isn't, this skill's own core already carries these rules — proceed without it rather than blocking.After producing or revising, hand it to the audit lenses rather than declaring it done. These are candidate lenses — posture per docs/orchestration-policy.md, or route the whole artifact through pm-product-review. Here `prioritization-rigor` is the always-relevant lens for this artifact (auto-runs) — a roadmap's integrity is its strategy-linked sequencing; the others are offered, tied to what the artifact actually contains. If the user invoked this skill for one specific thing, respect that scope.
- `pm-prioritization-rigor-audit` (auto-runs) — is the sequencing evidence-based, strategy-linked, and trade-off-explicit? Are scores real? Is anything on top because someone shouted?
- `pm-okr-metric-validity-audit` (offer — if it targets outcomes/KRs) — are the outcomes/KRs the roadmap targets themselves valid (falsifiable, non-vanity, guardrailed)? A roadmap aimed at bad metrics is well-organized waste.
If the audits surface a conflict (e.g., a stakeholder commitment vs. strategy fit), resolve back toward the primary task: the roadmap exists to align everyone on why and what outcome — protect its credibility over any single line item.
Common do/don't patterns
| ❌ Don't | ✅ Do |
|---|---|
| Features pinned to months, presented as promises | Now-Next-Later with explicit confidence levels |
| "Q3: Dashboard v2" | Theme → initiative → outcome, with the KR it serves |
| One roadmap document for everyone | Internal / executive / customer views of one truth |
| Roadmap items with story points and bug fixes | Roadmap = direction; backlog = tasks |
| Set-and-forget annual plan | Scheduled reviews + discovery-triggered replanning |
| Invent reach/impact numbers to look rigorous | Size from analytics; label estimates (→ prioritization-rigor) |
| Bury "why X isn't on it" | Explicit Won't-do/Later reasoning tied to strategy |
| Dense Gantt in the board deck | Short, visual, narrative presentation; details on request |
| Let sales-deal requests jump the queue silently | Structured intake, strategy filter, conscious exceptions |
| Ship without checking | Hand off to prioritization-rigor + okr-metric-validity |
