Install PM Skills Arsenal — GitHub Copilot

Load skill files into GitHub Copilot Chat or Copilot CLI to get elite-tier PM analysis. Same skills, same frameworks, same output quality as the Claude Code plugin.

Quick Start

Option A: Copilot Chat in VS Code / JetBrains

  1. Clone the repo into your workspace (or a folder your editor can see):

    git clone https://github.com/Avyayalaya/pm-skills-arsenal.git

  2. Open the folder in VS Code (or add it to your workspace).

  3. Reference a skill in Copilot Chat using #:

    #skills/competitive-market-analysis/SKILL.md

Our biggest competitor just made their core product free for teams
under 50. We're a Series B with 800 paying customers. Build a
competitive war map and recommend our response.

    Copilot loads the full skill file as context and produces framework-driven output.

Tip: You can also drag and drop the .md file directly into the Copilot Chat window, or use the Add Context button.

Option B: Copilot CLI

  1. Clone the repo somewhere accessible:

    git clone https://github.com/Avyayalaya/pm-skills-arsenal.git

  2. Reference the skill file in your prompt:

    #pm-skills-arsenal/skills/metric-design-experimentation/SKILL.md

We're launching real-time co-editing in our analytics product. 22K DAU,
W4 retention 52%. Build a measurement framework — NSM, experiment plan,
and retention cohort design.

Option C: No Clone (Copy-Paste)

  1. Open the skill file on GitHub:

  2. Click Raw, copy the full contents.

  3. Paste into Copilot Chat at the start of your conversation, then ask your question.

⚠️ What NOT to Do

Do NOT add skill files to .github/copilot-instructions.md or .github/instructions/.

Why:

  • Skill files are 1,000–1,100 lines each. GitHub recommends keeping instruction files under 1,000 lines — beyond that, Copilot processes instructions inconsistently.
  • Instruction files load on every Copilot request. A 1,000-line skill file would consume most of your context budget even when you’re doing unrelated work.
  • Skills are meant to be loaded on demand — reference them with # when you need them.

How It Works

When you reference a skill file with #, Copilot loads the full markdown content into its context window for that conversation. The skill contains:

  • Domain frameworks with scoring rubrics and decision tables
  • Application method — step-by-step with decision points
  • Quality gradients — Intern / Consultant / Elite tier definitions
  • Failure modes — what goes wrong and how to detect it
  • Mandatory output sections — Assumption Registry, Adversarial Self-Critique, Revision Triggers

The model reads the skill and applies the encoded frameworks to your question. No special commands or syntax needed — just ask naturally.

File Sizes & Context Budget

Skill Lines Size Fits in context?
Competitive & Market Analysis 1,099 91 KB ✅ Yes
Metric Design & Experimentation 1,050 93 KB ✅ Yes
Specification Writing 1,109 114 KB ✅ Yes

All three files fit comfortably within Copilot’s context window (200K+ tokens for current models). Loading one skill leaves ample room for your prompt, conversation history, and code context.

Tips for Best Results

  1. Load one skill per conversation. Each skill is self-contained. Loading multiple skills at once wastes context and can confuse framework application.

  2. Be specific in your prompt. The skill handles the frameworks — you provide the domain context. Include: your product, your market, your constraints, your data.

  3. Use the Quick Version for follow-ups. Each skill has a Quick Version (10–15 lines) inside the Full Version. After the initial analysis, you don’t need to re-reference the full file — the model retains the framework context.

  4. Check the Quality Gradient. Every skill output includes a self-assessment tier (Intern / Consultant / Elite). If output lands at Consultant tier, the skill’s Failure Modes section tells you what to fix.

Tested With

Model Status
Claude Sonnet 4.6 (via Copilot) ✅ Tested — full framework application
GPT-5.1 (via Copilot) ✅ Tested — strong output
GPT-5.2-Codex (via Copilot) ✅ Tested — strong output
Claude Haiku 4.5 (via Copilot) ✅ Tested — lower cost, good results

FAQ

Q: Does Copilot truncate the file if it’s too large? No. At 91–114 KB, skill files are well within Copilot’s context window. You’ll see the full framework applied in output.

Q: Can I use this with GitHub Copilot in the browser (github.com)? Yes. Use the copy-paste method — paste the raw skill file contents into the Copilot Chat on github.com, then ask your question.

Q: Can I reference multiple skill files at once? Technically yes, but not recommended. Each skill is 1,000+ lines. Loading two would consume significant context and dilute framework application.

Q: How is this different from custom instructions? Custom instructions (.github/copilot-instructions.md) are short, persistent rules for every request. Skills are deep domain knowledge loaded on demand for specific tasks. They serve different purposes.

Back to Home · GitHub Repo · MIT License