How the /discover Command Chains Skills for Product Discovery

The /discover command in the phuryn/pm-skills repository orchestrates four core skills—brainstorm‑ideas, identify‑assumptions, prioritize‑assumptions, and brainstorm‑experiments—into a sequential workflow that generates a complete discovery plan from initial context to validated experiments.

The /discover command serves as the primary entry point for product managers using the phuryn/pm-skills framework. It transforms fragmented discovery activities into a structured, reusable pipeline by automatically routing inputs through specialized skills based on product stage. This command demonstrates how modular AI skills can be composed to handle complex, multi-stage product discovery workflows without manual context switching.

Step-by-Step Skill Orchestration

The command implementation in pm-product-discovery/commands/discover.md defines a seven-stage pipeline that conditionally routes data between stage-specific skill variants.

Context Capture and Stage Routing

The workflow begins by querying the user to determine whether the product is existing or new, then collects supporting artifacts such as research documents, PRDs, or interview transcripts. This routing logic sets the product_stage variable that determines which skill variants are invoked in subsequent steps.

Divergent Idea Generation

Depending on the stage flag, the command invokes either brainstorm-ideas-existing or brainstorm-ideas-new. These skills, documented in pm-product-discovery/skills/brainstorm-ideas-existing/SKILL.md and pm-product-discovery/skills/brainstorm-ideas-new/SKILL.md, generate ten candidate ideas from the perspectives of Product Manager, Designer, and Engineer. The user selects promising candidates to feed into the next stage.

Assumption Surface and Categorization

For each selected idea, the workflow calls identify-assumptions-existing or identify-assumptions-new. According to the source files pm-product-discovery/skills/identify-assumptions-existing/SKILL.md and pm-product-discovery/skills/identify-assumptions-new/SKILL.md, these skills extract risky assumptions across Value, Usability, Feasibility, and Viability categories. For new products, they additionally surface Go-to-Market, Strategy, and Team risks.

Impact-Risk Prioritization

The aggregated assumptions are passed to the prioritize-assumptions skill, implemented in pm-product-discovery/skills/prioritize-assumptions/SKILL.md. This skill maps assumptions on an Impact × Risk matrix, flags leap-of-faith assumptions, and returns a ranked list to guide validation efforts.

Experiment Design and Sequencing

The top-ranked assumptions flow into brainstorm-experiments-existing or brainstorm-experiments-new. As defined in pm-product-discovery/skills/brainstorm-experiments-existing/SKILL.md and pm-product-discovery/skills/brainstorm-experiments-new/SKILL.md, these skills propose concrete validation methods—such as A/B tests, fake doors, landing-page pre-launches, and concierge MVPs—complete with success criteria, effort estimates, and execution sequencing.

Discovery Plan Assembly

All outputs are compiled into a markdown discovery-plan document that records the selected ideas, assumptions, experiment matrix, timeline, and decision framework. The plan is saved to the user’s workspace as a file such as discovery-smart-notif.md.

Follow-Up Action Routing

Finally, the command presents optional next steps—including generating a PRD, drafting interview scripts, setting up metrics, or creating user-story estimates—allowing the user to extend the workflow without leaving the conversational context.

Example Invocation and Flow

The following demonstrates a typical interaction with the chained skill system:

# Example invocation

/discover Smart notification system for our project management tool

# Typical interaction flow

User: /discover Smart notification system for our project management tool

Bot: 1️⃣ What stage is this product? (existing / new)  
Bot: 2️⃣ Here are 10 brainstormed ideas … (from brainstorm-ideas-existing)  
User: I’ll keep ideas 1, 3, 5.  

Bot: 3️⃣ Identifying assumptions for those ideas … (identify-assumptions-existing)  
Bot: 4️⃣ Prioritizing assumptions … (prioritize-assumptions)  
Bot: 5️⃣ Designing experiments … (brainstorm-experiments-existing)  

Bot: 📄 Discovery plan generated – saved as `discovery-smart-notif.md`.  

Bot: What next? (create PRD, draft interview script, set up metrics, …)

Technical Architecture of the Skill Chain

The chaining mechanism relies on conditional routing based on the product_stage variable. Each skill accepts structured input from the previous stage and returns standardized output for the next, creating a pipeline where context persists across invocations.

The command handler manages state transitions between the six primary skills, ensuring that outputs from brainstorm-ideas feed into identify-assumptions, and that prioritized assumptions automatically trigger the appropriate brainstorm-experiments variant. This architecture prevents context loss between discrete discovery activities.

Summary

• The /discover command acts as an orchestrator for four core skills: brainstorm-ideas, identify-assumptions, prioritize-assumptions, and brainstorm-experiments.
• It routes execution to stage-specific variants (existing vs. new) based on initial user input captured in pm-product-discovery/commands/discover.md.
• The workflow generates a markdown discovery plan documenting ideas, assumptions, and experiments with specific file outputs.
• Source implementation files are located in pm-product-discovery/skills/ with each skill defined in its respective SKILL.md file.

Frequently Asked Questions

How does the /discover command choose between existing and new product skills?

The command prompts the user to specify the product stage during the Context Capture phase. If the user selects "existing", the workflow invokes brainstorm-ideas-existing, identify-assumptions-existing, and brainstorm-experiments-existing. For "new" products, it switches to the corresponding *-new skill variants defined in the repository.

What output format does the /discover command generate?

The command produces a markdown discovery plan saved to the workspace (e.g., discovery-smart-notif.md). This document contains the selected ideas, categorized assumptions, prioritized Impact-Risk matrix, and the experiment design with success criteria and effort estimates.

Can I skip specific steps in the discovery chain?

The current implementation in phuryn/pm-skills processes the full chain sequentially to ensure completeness. While you cannot skip core steps within a single /discover invocation, you can invoke individual skills directly (e.g., /skill prioritize-assumptions) if you need to execute only one specific phase with manual inputs.

What types of experiments does the brainstorm-experiments skill suggest?

The skills suggest validation methods including A/B tests, fake door tests, landing-page pre-launches, concierge MVPs, and usability studies. Each proposal includes specific success criteria, effort estimates, and sequencing recommendations tailored to the assumption being validated, as specified in pm-product-discovery/skills/brainstorm-experiments-existing/SKILL.md and pm-product-discovery/skills/brainstorm-experiments-new/SKILL.md.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}