Practical Agent Skills for Product Teams

🧠 McKinsey-style charts, benchmark competitor analysis, translate specs into backlog items, single page animated HTML presentations. 10 skills you can use at work.

🔒 The Knowledge Series breaks down emerging AI technologies with practical playbooks designed specifically for product teams. Get 100+ guides and practical tutorials covering everything from Claude Code and MCP to agentic workflows, vibe coding, and more.

---

After being created less than a year ago, agent skills are fast becoming an essential part of the product development process at some of the world’s leading companies.⁠

Anthropic recently revealed that their automated reporting tool uses agent skills to boost reporting accuracy from 21% to 95%, Pinterest uses them for automated testing and fintech company Ramp values agent skills so much that they built their own internal library of shareable skills in a custom-built tool called Glass.

Perplexity estimates that there are now in the region of 40,000 to 100,000 agent skills published across various code repositories with private repos likely pushing that number well over the 100,000 mark. Needless to say, with so many skills now published, navigating them all to find the ones that are actually useful is becoming increasingly difficult.

Coming up:

In this Knowledge Series, we’ll share 10 hand-picked agent skills specifically designed for product teams. These 10 skills have been crafted by people at companies including Google and YCombinator and will allow you to:

1. Refine ideas before you ship them using product frameworks

2. Generate McKinsey-style consulting charts you can edit - and then critique

3. Benchmark competitors across nine weighted dimensions like positioning,visual craft and strategic tension

4. Create design systems from scratch with a design consultant skill

5. Build pricing pages based on your DESIGNmd files

6. Create structured plans before starting any multi-step work

7. Brainstorm ideas through natural, collaborative dialogue

8. Send internal company updates including company newsletters, status reports and leadership updates

9. Design beautiful single page animated HTML presentations

10. Translate specs into backlog items

---

The Knowledge Series

What are agent skills again? A quick recap

Skills are folders containing markdown files and scripts that an LLM loads when it decides they’re relevant to a task.

The key file is SKILL.md, which has three layers: a lightweight header describing what the skill does (so the LLM can quickly decide if it’s needed), a detailed body with step by step instructions, and file references for scripts or assets that only get pulled in when actually required. This layered approach is called progressive disclosure - it keeps Claude and other tools from loading everything at once and wasting context.

Read the full guide on Agent Skills in a previous Knowledge Series here.

Skills can also stack together to handle complex tasks. So, for example, when you give Claude a prompt, it scans available skills, picks the ones that fit, adds them to context, runs any scripts they include, and then completes the task using that specialized guidance.

You can explicitly mention a skill if you want to but no special invocation is needed - it happens automatically based on what you ask for. Anthropic ships some pre-built skills (PowerPoint, Excel, Word, PDF) that work without being toggled on, and you can also build or import custom ones, like a brand guidelines skill or Notion’s spec-to-implementation skill.

10 practical agent skills you can use

Here’s a collection of 10 practical agent skills you can use. We’ll explore how the skills work with some real world scenarios and you can also download a folder containing each of hand-picked skills to make it easier to install in your own set up at the end of this post.

1. Refine and stress test new ideas before you ship them

Repo: github.com/addyosmani/agent-skills

This skill was developed by Google’s Addy Osmani and is designed to take a rough product idea and work it into something you might actually want to ship and is ideal for scenarios like design workshops.

It uses the “how might we” open ended questions that Google’s product teams are known for at this stage. Other trigger phrases include:

• “Help me refine this idea”

• “Ideate on [concept]”

• “Stress-test my plan”

It first pushes the idea outward, asking questions and generating variations, then pulls everything back in, grouping the strongest options and poking holes in them to try and find all of the assumptions you didn’t realize you were making.

It comes with several product ideation frameworks built directly into it. These frameworks include

• SCAMPER - Seven prompts (Substitute, Combine, Adapt, Modify, Put to other uses, Eliminate, Reverse) that push you to transform an existing product or feature. Works best when you already have something to improve, not for starting from scratch.

• How Might We (HMW) - Turns a pain point into a question shaped like “How might we [outcome] for [user] without [constraint]?” Useful for pulling someone away from a solution they’re stuck on and back to the actual problem.

• First Principles Thinking - Strips an idea down to what’s actually true, separates that from assumption, then rebuilds from scratch. Good for breaking out of small incremental tweaks when everything feels like a minor variation on what already exists.

• Jobs to Be Done (JTBD) - Focuses on the task the user is hiring a product to do, split into functional, emotional, and social jobs. Helps you check whether you’re solving the right problem in the first place.

• Constraint-Based Ideation - Forces creativity by imposing artificial limits, like one day to build it, one feature only, or free forever. Good for cutting an idea down when it’s gotten too big or too vague.

• Pre-mortem - Imagines the project has already failed a year from now and works backward to find every plausible cause. Best used later, once an idea already feels good, to stress-test it before committing.

• Analogous Inspiration - Looks at how other industries, companies, or even nature solved a structurally similar problem, not just a surface-level one. Good early on for generating ideas that feel genuinely different rather than obvious variations.

Example

In this example, you can see how the skill could be used to explore a feature idea within an existing product using the built-in frameworks:

If you’re running design or strategy workshops about what types of new features to ship next, the idea-refine skill could be one to tap into.

2. Generate McKinsey-style consulting charts you can edit - and then critique your deck

Repo: github.com/sruthir28/enterprise-ai-skills

This one builds you consulting-style charts for a deck and is designed to be a clean, clutter-free alternative to some of the higher density charts that consultants can be known for. The charts it produces are natively editable rather than just static image files and will work inside Powerpoint or Keynote (which you can then upload to Google Slides if necessary).

The skill has some design opinions of its own to make the format easily digestible, so for example, it limits colors to use grey and with accent colors to make the data stand out:

Once installed, you could invoke it using a prompt like this:

“Generate a McKinsey-style bar chart showing our TAM growing from $24.8B in 2023 to $42.5B in 2028, highlight the final year, and add a callout with the 2028 number.”