How to Use pm-skills with AI Assistants Other Than Claude

Yes, pm-skills works with Gemini CLI, OpenCode, Cursor, Kiro, and any AI assistant that can read markdown files.

The pm-skills repository by phuryn separates framework knowledge from execution logic, making it portable across different AI platforms. While Claude uses the full feature set including slash commands, other assistants can leverage the same product management frameworks by accessing the plain-text markdown skills stored in each plugin's directory.

Understanding the pm-skills Architecture

The repository organizes content into two distinct layers that determine cross-platform compatibility.

Skills (Assistant-Agnostic)

Skills are plain-text markdown files stored under each plugin's skills/ folder, such as pm-product-discovery/skills/brainstorm-ideas-new/SKILL.md or pm-product-strategy/skills/product-vision/SKILL.md. These files contain the complete prompts and frameworks that describe product management methodologies. Because they are standard markdown without proprietary formatting, any AI assistant can read and execute them.

Commands (Claude-Specific)

Commands reside in commands/ folders like pm-product-discovery/commands/discover.md and provide Claude-specific slash-command definitions that orchestrate multi-step workflows. These are the only Claude-dependent components and can be omitted when using other assistants.

Setting Up pm-skills for Other AI Assistants

To use pm-skills with assistants other than Claude, copy the skill files into your target platform's designated skills directory.

Supported Installation Paths

Different assistants expect skills in specific locations:

• Gemini CLI: ~/.gemini/skills/ (global) or .gemini/skills/ (project-level)
• OpenCode: .opencode/skills/ (project-level)
• Cursor: .cursor/skills/ (project-level)
• Kiro: .kiro/skills/ (project-level)

Automated Copy Script

Run this shell command from the repository root to copy all skills to your chosen assistant's directory. This script is adapted from the implementation in README.md:

# For OpenCode (project-level)

for plugin in pm-*/; do
  mkdir -p .opencode/skills/
  cp -r "$plugin/skills/"* .opencode/skills/ 2>/dev/null
done

For Gemini CLI global installation, modify the destination path:

for plugin in pm-*/; do
  cp -r "$plugin/skills/"* ~/.gemini/skills/ 2>/dev/null
done

Using Skills with Different Assistants

Once copied, invoke skills using your assistant's specific command syntax while referencing the skill name.

Gemini CLI Example

gemini skill brainstorm-ideas-new \
  "Generate 5 high-impact ideas for a new B2B analytics platform."

OpenCode Example

opencode skill product-vision \
  "Create a compelling vision for an AI-driven project-management SaaS."

Since the skill file contains the full prompt framework, the assistant returns structured guidance based on the markdown content without requiring any translation or modification.

Recreating Claude Commands for Other Platforms

If you need the automated workflow orchestration that Claude commands provide, you must recreate the logic in your target assistant's scripting format. The core skill calls remain identical.

For example, the discover command in pm-product-discovery/commands/discover.md chains multiple skills together. You can recreate this in Python:

def discover_workflow(idea):
    ideas = opencode.skill("brainstorm-ideas-new", idea)
    assumptions = opencode.skill("identify-assumptions-new", ideas)
    prioritized = opencode.skill("prioritize-assumptions", assumptions)
    experiments = opencode.skill("brainstorm-experiments-new", prioritized)
    return {
        "ideas": ideas,
        "assumptions": assumptions,
        "prioritized": priorities,
        "experiments": experiments
    }

# Execute the workflow

result = discover_workflow("AI-powered meeting summarizer")

This approach preserves the pm-skills framework logic while adapting the execution layer to your assistant's capabilities.

Summary

• pm-skills stores product management frameworks as plain markdown in skills/ directories, making them universally readable.
• Claude-specific commands in commands/ folders are optional and can be replaced with your assistant's native scripting.
• Gemini CLI, OpenCode, Cursor, and Kiro are explicitly supported through standard directory copying.
• No file modification is required to use skills across different platforms.
• Workflow automation requires recreating command logic, but the underlying skill calls remain unchanged.

Frequently Asked Questions

Can I use pm-skills with ChatGPT or other web-based AI assistants?

Yes, but with manual steps. Since ChatGPT lacks a native skills directory, you must copy the content from skill files like pm-product-discovery/skills/brainstorm-ideas-new/SKILL.md and paste it into your conversation as a system prompt or custom instruction. The frameworks execute correctly, though you lose the automated command chaining available in Claude.

Do I need to modify the skill files to use them with other assistants?

No. The skills are standard markdown files that any AI assistant can parse. The repository deliberately stores prompts in plain text to ensure portability across platforms including Gemini CLI, OpenCode, Cursor, and Kiro.

What is the difference between skills and commands in pm-skills?

Skills are markdown files containing the actual product management frameworks and prompts, stored in paths like pm-execution/skills/prd-writing/SKILL.md. Commands are Claude-specific orchestration files that chain multiple skills together into automated workflows, located in commands/ directories. Skills are universal; commands are Claude-only.

Is there a performance difference when using pm-skills with different AI assistants?

Performance depends on the underlying model rather than the pm-skills format. Since skills are plain markdown, the execution speed is determined by your assistant's API latency and token processing capabilities. Complex skills like pm-product-strategy/skills/product-vision/SKILL.md may generate longer responses, but this behavior is consistent across platforms given the same model.

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