Best Claude Code Skills for Product Managers in 2026

Product management is the role with the worst tool sprawl. You write in Notion, plan in Linear, present in Figma, model in Sheets, sync in Slack, and spend forty percent of your week shuffling context between them. Agent skills running in Claude Code can collapse most of that overhead. If you pick the right ones.

This is a working list of seven skills product managers actually install and keep installed. Each has a GitHub source, a security caveat where relevant, and an install command. For the broader profession landing page, see our full list for product managers.

What an agent skill is, in PM terms

An agent skill is a folder of instructions plus optional scripts that teaches Claude Code how to do a specific job. Different from MCP servers (which are network services). Skills run locally as part of your Claude Code session.

You install one by running the /plugin install slash command inside Claude Code, or by cloning the GitHub repo into ~/.claude/skills/. The skill loads when its description matches your prompt. So if a skill says "use when writing PRDs," it activates when you ask Claude to draft a PRD.

For a fuller primer on the install mechanics, see how to install the frontend-design skill. Same pattern applies to every skill below.

Quick comparison table

1. product-manager-skills. The PM toolkit

A bundled skill that covers the core PM artifacts: PRDs, one-pagers, OKR drafts, sprint goals, customer interview synthesis. Less individual depth than specialized skills, more breadth. Useful if you do not want to install five separate skills for PM-specific tasks.

The PRD template alone is worth the install. It enforces the structure most teams already use (problem, target user, success metrics, scope, out-of-scope, risks) and pushes back when you skip a section. New PMs especially benefit from this. The skill works as a junior coach checking your spec for the obvious gaps.

Install: git clone into ~/.claude/skills/product-manager-skills/

Security: Community skill. Read the SKILL.md file before installing. It is a long prompt, no scripts, but verify nothing is doing remote calls you did not expect.

2. notion-mcp. Notion as a Claude data source

Notion is where most PM documentation lives. The notion-mcp skill connects Claude Code to your Notion workspace via OAuth, so Claude can read pages, query databases, and write back drafts.

Realistic workflow: you have an old PRD in Notion for a feature being revisited. Ask Claude "summarize the existing PRD for feature X, then propose updates based on the new customer requirements I am about to paste." Claude reads the page, ingests your new input, and drafts a revision into a new Notion subpage.

Install: Use the Notion MCP integration through Claude Code's plugin marketplace, or follow the setup at notion.so/help/mcp.

Security: OAuth-scoped tokens. Grant access only to the workspace and pages Claude needs. Notion's permission model is fine-grained enough that you can restrict the integration to a single PM project space.

3. linear-mcp. Linear as a Claude data source

Linear's MCP server does the same job for Linear issues. Claude can read your team's sprint, summarize issue status, draft new issues from a meeting transcript, and update existing issues with new comments.

Where this saves real time: standup prep. Ask Claude "summarize what changed in our Linear board since yesterday at 9am, group by status and assignee, flag anything blocked." You get a one-paragraph standup script instead of clicking through 30 issues yourself.

Install: Linear MCP server from Linear's developer docs, plus an API key from your Linear settings.

Security: API key with the minimum scopes needed (issues:read, issues:write if you want creation). Do not grant admin scopes.

4. deep-research. Customer research synthesis

The deep-research skill is Anthropic's tool for the multi-step research workflow: gather sources, synthesize across them, produce a structured summary with citations. PMs use it for competitive analysis and customer-interview synthesis.

Practical example: you have got 12 customer interview transcripts from the last month. Ask Claude with deep-research active to "synthesize the themes across these interviews, weight by how many customers mentioned each, surface the top three job-to-be-done patterns with quoted evidence." The output is a research memo you can drop into your next product review.

Install: Available in Claude Code's plugin marketplace.

Security: The skill does web fetches when researching external sources, so be aware of what URLs it is hitting. For internal interview transcripts that stay local, no network exposure.

5. xlsx-spreadsheet. Roadmap models and RICE scoring

PMs end up in spreadsheets for prioritization, capacity planning, and any model that has more than three variables. The xlsx-spreadsheet skill lets Claude read and write Excel files directly, which means you can describe the model you want and Claude builds the spreadsheet.

Realistic ask: "Build me a RICE scoring sheet for our Q3 backlog. Pull the items from this CSV, add Reach/Impact/Confidence/Effort columns with formula-driven priority scores, and sort by descending score." You get a working .xlsx file in 30 seconds.

Install: Available in Claude Code's plugin marketplace as xlsx-spreadsheet.

Security: Local file ops only. The skill reads and writes .xlsx files on your machine. No network calls.

6. claude-seo. Landing page + content briefs

If your PM scope includes the marketing site or any content surface, claude-seo gives you the same SEO research, content brief, and on-page audit workflow used by marketing teams. For a PM owning growth, this means you can hand a structured content brief to the marketing team rather than vague requirements.

Where it earns its slot: feature launch landing pages. You can use the skill to generate a keyword-optimized brief that the writer fills in, then audit the published page against on-page SEO criteria before it goes live.

Install: Available in Claude Code's plugin marketplace.

Security: Community skill. Audit the prompt file before installing. The skill does WebFetch calls when researching SERPs, so confirm you are comfortable with that. See our claude-seo review for the deeper audit.

7. doc-coauthoring. For specs that turn into documents

For PMs who write long-form artifacts (a PRD that ends up 20+ pages, a strategy doc, an internal whitepaper), the doc-coauthoring skill helps Claude maintain coherence across a long document. It handles cross-references, consistent terminology, and structural revisions without losing track of what is already been written.

The skill shines when you are past the outline stage and refining. Ask Claude to "tighten section 3.2 of this spec without changing the technical claims, and verify any cross-references in sections 4 and 5 still resolve." That kind of long-document editing is where most general-purpose AI breaks down.

Install: Available in Claude Code's plugin marketplace.

Security: Local file ops only.

How to install any of these

Inside Claude Code, run:

/plugin marketplace add anthropics/skills
/plugin install <skill-name>@<source>

For community skills outside the official marketplace, clone the GitHub repo into ~/.claude/skills/<skill-name>/. The skill loads automatically when Claude Code restarts.

Before installing any community skill, read the SKILL.md file. It is the actual prompt the skill injects into Claude. If anything looks off — references to external URLs you do not recognize, scripts that do not match what the description claims — do not install it.

Two skills not on this list, and why

JIRA-specific skills. Several exist, but the official linear-mcp and notion-mcp patterns are cleaner and the JIRA ecosystem options have not matured to the same level. If your team is on JIRA, the workflow tends to be "export to CSV, use xlsx-spreadsheet" until JIRA ships a first-party MCP.

AI-prioritization skills that auto-score your backlog. These exist and the demos look great. In practice, the priority a PM ships is not just RICE math. It is RICE math weighted by partnership commitments, internal politics, and what your VP said in last Thursday's leadership meeting. AI can build the model. The PM still decides what goes in Q3.

Stack recommendation by PM seniority

Junior PM (first 1-2 years): product-manager-skills + notion-mcp. Get the PRD muscle memory and reduce the time you spend reformatting docs.

Mid-level PM (3-5 years): Add linear-mcp + xlsx-spreadsheet + deep-research. You are managing roadmaps, prioritization, and customer research at a heavier load. Automate the synthesis layer.

Senior PM / Lead PM: Add claude-seo + doc-coauthoring. Strategy docs, cross-functional content, and the growth-adjacent work all benefit from these.

FAQ

Are these skills free?

Yes. The skills themselves are open-source or part of Anthropic's free skill marketplace. You pay for Claude Code (Pro at $20/mo or Team at $25/seat). Some skills connect to paid services like Notion or Linear. Those costs are separate.

Will my company's IT block these?

Notion-mcp and linear-mcp require OAuth or API keys, which means IT has visibility into the integration. The local-only skills (xlsx-spreadsheet, doc-coauthoring) do not touch the network. Talk to IT before installing the ones that connect to your work accounts.

Can I use these without Claude Code?

The skills are designed for Claude Code specifically. The underlying capabilities exist in the Claude API, but you would be building the orchestration layer yourself.

What about MCP servers for Slack or Figma?

Slack MCP and Figma MCP both exist and are useful but neither is mature enough yet to recommend as a PM staple. Watch this space. Both are likely to become standard PM tools within 6 months.

Should I write my own skill?

For repetitive tasks specific to your team's workflow — your team's status report format, your roadmap review template — yes, custom skills pay back fast. The skill-creator skill walks you through it. Start with a 50-line SKILL.md and grow it.

What is the biggest mistake PMs make with these?

Treating Claude like a deliverable factory. The skills are best when they accelerate your thinking. Drafts you react to, models you tweak, syntheses you challenge. PMs who hand off the strategic call to AI end up with generic roadmaps and lose the muscle that justifies their job.