Dean Peters
6a4fbf20c8
Every skill now declares its inputs in a plain-language ## Input section
(Purpose > Input > Key Concepts...): what it works best with, what else
sharpens the output, the inline-input rule (context supplied with the
invocation counts as answers already given — used, not re-asked), an
empty-handed fallback ("invitation, not gate"), and an example invocation.
- Enforce the convention: check-skill-metadata.py requires the Input
section and fails bare $ARGUMENTS in skill bodies; smoke checks warn
on missing example invocations or empty-handed language
- Document the differentiator: CONTRIBUTING.md "Why We Don't Use
$ARGUMENTS" (portability, pedagogy, unnecessary), with short versions
in README, AGENTS.md, and Building PM Skills
- Add argument-hint frontmatter to 53 skills for Claude Code autocomplete
- Streamlit: render each skill's Input section in a "What to bring
(all optional)" pre-flight expander
- Add workshop-facilitation/examples/inline-input-flow.md — transcript
of inline input skipping answered questions, plus the re-asking
anti-pattern
- Restore skills/agent-orchestration-advisor (Phase 6 skill #34): it
only ever existed on orphaned commit a41415c while docs claimed it
shipped; recovered and upgraded to current standards
- Sync .claude-plugin/marketplace.json (47 -> 55 entries) and bump
plugin manifests to 0.81
- Update anatomy and release docs: CLAUDE.md, CONTRIBUTING.md, AGENTS.md,
README.md, usage guides, authoring scripts/adapters, CHANGELOG,
announcement, catalogs
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
3.1 KiB
Using PM Skills with ChatGPT
You can use this skills repo with ChatGPT in three ways, depending on how repeatable you want the setup.
If you are new to this repo, start with Using PM Skills 101.md.
If you specifically use the desktop app, see the PM-first guide: Using PM Skills with ChatGPT Desktop.md.
If you want to test a skill locally first, use the Streamlit (beta) playground:
pip install -r app/requirements.txt
streamlit run app/main.py
Option 1: Connect GitHub in ChatGPT Apps (Best for Active Repos)
Use this when you want ChatGPT to read current files from GitHub.
Availability can vary by plan and rollout region.
- In ChatGPT, open Settings.
- Go to Apps (or Connected Apps, depending on UI version).
- Connect GitHub and authorize access.
- In a chat, ask ChatGPT to use a skill file from this repo by path.
Example:
Use skills/prioritization-advisor/SKILL.md from deanpeters/Product-Manager-Skills and guide me through choosing a framework for a B2B roadmap.
Command-style example:
Run commands/strategy.md from deanpeters/Product-Manager-Skills for: B2B analytics add-on for mid-market ecommerce brands.
Option 2: Build a Custom GPT with Skill Knowledge (Best for Reuse)
Use this when you want a reusable PM assistant that always carries selected skills.
- In ChatGPT, create a new GPT.
- Upload skill files as Knowledge.
- Add instructions telling the GPT to follow the skill structure exactly.
- Save and test.
Important for this repo:
- ChatGPT GPT Knowledge expects uploaded files; no GitHub auto-sync for these uploaded files.
- This repo stores source files as
SKILL.md. For Claude-style compatibility packaging, run:
bash scripts/package-claude-skills.sh
Then upload the generated files from dist/claude-skills/<skill-name>/Skill.md (or upload the original SKILL.md files directly if you are only using ChatGPT).
Option 3: Use ChatGPT Projects with Files (Best for One Initiative)
Use this for project-scoped work where you want files, chats, and instructions grouped together.
- Create a Project in ChatGPT.
- Upload one or more skill files (or a small curated set).
- Add project instructions that require skill-conformant output.
- Run your PM tasks inside that project.
Recommended Project/GPT Instruction Snippet
When solving PM tasks, use the uploaded skill files as the operating standard.
Follow each skill's sections in order: Purpose, Input, Key Concepts, Application, Examples, Common Pitfalls, References.
If context is missing, ask up to 3 clarifying questions before drafting.