How to use the Product And Business Analysis agent skill

What is an agent skill and what does this one do?

An agent skill is a folder of instructions, centred on a SKILL.md file, that an AI agent loads when a task matches the skill's description. This skill's description says it applies when a decision depends on metric-backed evidence, such as choosing a direction, prioritizing an opportunity, evaluating a change, segmenting users, sizing tradeoffs, or deciding what to do next.

When it runs, the skill makes the agent work in a fixed order: first pin down the decision and who will act on it, then gather relevant context, frame specific data questions, run focused quantitative checks against proper baselines, and finish with a recommendation that states its uncertainty. It also tells the agent to check how reliable each data source is and to reconcile conflicting sources before drawing conclusions.

• Starts from the decision, audience and action rather than from whatever data is easiest to pull.
• Compares results to a relevant baseline or denominator, for example checking whether high usage just reflects a larger group.
• Sizes the opportunities it finds and keeps unclassified or unknown groups visible when they could change the answer.
• Recommends validating numbers before sharing high-impact or surprising results, and reconciling dashboards with direct queries when both exist.

Install the skill from the ZIP download

Download the ZIP for the skill from our page and unzip it. The skill's folder contains agents/openai.yaml and SKILL.md. Place the unzipped skill folder in one of two locations, depending on how you want to use it:

For all your projects, put it in ~/.claude/skills/. For a single project, put it in that project's .claude/skills/ folder. Restart or start a new session in your agent tool so it picks up the new skill.

• Personal install: ~/.claude/skills/product-business-analysis
• Project install: .claude/skills/product-business-analysis inside your project folder
• Claude apps: upload the skill under Settings → Capabilities → Skills instead of copying files.

Install the companion skills it depends on

This skill refers to several other skills by name and relies on them for parts of the work. The companion skills are gather-business-context, design-kpis, metric-diagnostics, jupyter-notebooks, analyze-data-quality and validate-data. The source does not supply installation files for these, so you need to obtain and install them separately.

Without the companions the core workflow still exists, but steps such as context gathering, KPI definition, data quality checks and validation will not be fully covered.

Connect the data sources it needs

The skill expects live reads from your data. It works best when the agent can reach your warehouse, dashboards or semantic layer. The source does not name specific products, so the connection method depends on your own setup.

The skill is built to stop rather than guess. If a source it needs for the answer is unavailable, it tells you which source is missing and asks you to make it available or to approve a reviewed fallback. It does not treat a weaker substitute as equivalent.

• Give it read access to the warehouse, dashboards or semantic layer that hold the metrics you care about.
• Tell it which source is the source of truth if you know it, to avoid conflicts between sources.
• If you lack data access, the skill will stop on any path that needs it, so it will not produce a metric-backed answer from memory.

Trigger the skill with a natural request

You do not need special syntax. Describe a decision in plain language and include the decision context, the audience and the comparison you care about. The skill's description lists the kinds of requests it matches, so phrase your request around a choice rather than a single number.

The skill also refers to itself with a $ prefix in its instructions, for example $gather-business-context, which names the companion skill it calls. You can ask for the analysis directly in your own words.

• "Should we prioritize the team invite feature or the reporting export? Use our usage and 30-day retention data to recommend one, and state your assumptions."
• "Segment our trial users by how quickly they reach their first key action and tell me which segment is worth targeting next quarter."
• "Evaluate whether the new onboarding flow should roll out to all regions, using the conversion data and flag any conflicts between dashboards and the warehouse."

Know when a different skill is a better fit

The skill is designed for decision-level analysis. The source says it is heavyweight for quick one-off lookups or simple metric questions, so ask a plain question directly for those.

If the question is why a metric moved, the source says that is better handled by the metric-diagnostics skill. The product-business-analysis skill points to it when the recommendation depends on explaining a metric movement, anomaly, gap or discrepancy.

• Simple lookup: ask the agent directly without this skill.
• Unexpected metric change on its own: use metric-diagnostics.
• Decision on which direction to take: use this skill.

Safety notes on scripts and network access

According to the source, this skill does not run scripts. It does use the network, because it reads from live data sources. Review which connections you grant the agent, and make sure they are read access to the data you intend it to analyse.

The source does not describe any data retention or privacy behaviour, so check your data policies before connecting sensitive sources.

• Runs scripts: no.
• Uses network: yes, for live reads of your data sources.
• Requires access to user data sources; the source does not name specific products.