How JD Health’s OPC Team Built a Full‑Process Product Skill Framework

Background and Motivation

JD Health’s One‑Person Company (OPC) teams lack a dedicated product role, so they must answer questions such as whether a problem is worth solving, what evidence supports it, if the current solution is unique, and how to measure success after launch.

Adopted Framework

The team leverages Anthropic’s open‑source product‑management plugin , which provides eight reusable Skills that break the product lifecycle into discrete, traceable actions.

Installation of Skills

https://github.com/anthropics/knowledge-work-plugins/tree/main/product-management/skills

Install the eight skills (synthesize‑research, metrics‑review, competitive‑brief, product‑brainstorming, write‑spec, roadmap‑update, sprint‑planning, stakeholder‑update) in JoyCode or Codex by entering the provided prompt.

Decision‑Making Chain

Problem Judgment – assess if the issue is worth solving.

Requirement Definition – turn the direction into actionable specs.

Roadmap Execution – place the requirement into priority, capacity, and dependency constraints.

Launch Review – validate outcomes with metrics and share results.

Every demand must pass the six checkpoints: evidence, hypothesis, boundary, scheduling, acceptance, and review.

Stage 1: Problem Judgment

Use the synthesize-research skill with inputs (direct paste, file upload, tool connections) formatted with five fields: source, time, object, content, and relation. The goal is to decide importance, not to prove a pre‑chosen solution.

Stage 2: Requirement Definition

When a problem is validated, run product‑brainstorming to explore problem space, solution space, and key assumptions, then generate a structured spec with sections for Problem, Evidence, Goals, Non‑goals, Requirements (P0‑P2), Acceptance Criteria, Metrics, and Open Questions.

Stage 3: Roadmap Execution

Use roadmap‑update to decide placement (Now/Next/Later), alignment with OKRs, dependencies, and trade‑offs. Follow with sprint‑planning to set sprint goals, capacity, carry‑over reasons, priority tiers, owners, risks, and Definition of Done, reserving 20‑30% capacity for unplanned work.

Stage 4: Launch Review

After release, run metrics‑review to compare actual metrics against the original goals, identify leading/lagging signals, user impact, any negative effects, and decide next steps (iterate, expand, roll back, or stop).

Stakeholder Synchronization

Use stakeholder‑update to tailor communication for different audiences, employing status colors (Green, Yellow, Red) to highlight risk levels and drive early mitigation.

Quick Start Guide

Install the eight skills (≈10 min).

Select a small, real‑world demand.

Run three core skills: synthesize-research, write-spec, metrics-review.

Gradually add product‑brainstorming, roadmap‑update, sprint‑planning, and stakeholder‑update.

Document templates (material input format, spec, roadmap change log, review checklist) for future reuse.

Common Pitfalls

Skipping problem judgment and jumping straight to PRD.

Using Skills to justify a pre‑selected solution.

Feeding raw material without source, time, object, and hypothesis fields.

Treating new demand as “extra work” instead of a capacity reallocation.

Planning 100 % capacity, leaving no buffer for incidents.

Concluding a post‑launch review with only “launched” without metric analysis.

Using the same report for all audiences instead of tailoring content.

Conclusion

The greatest value of the Skill set is not the AI tools themselves but the shared responsibility it creates for product and engineering teams, turning each decision into a documented, traceable, and repeatable asset.