Three of the agent "skills" Wade Foster (Co-Founder & CEO, Zapier) demoed during his Running Remote 2026 fireside chat, cleaned up into general-purpose versions you can drop into your own AI setup.
A "skill" here is just a Markdown file that tells an AI agent (Cursor, Claude Code, Claude Projects, a custom GPT, etc.) how to run a specific workflow. Each one is written to be tool-agnostic: wherever the original relied on a specific tool, these versions use a generic placeholder like [YOUR MEETING NOTES TOOL] that you fill in once.
| Skill | What it does |
|---|---|
war-council/SKILL.md |
Convenes a panel of opinionated expert personas (ruthless CFO, contrarian board member, customer obsessive, plus experts generated for your specific problem) to stress-test a decision and return a ranked, bet-weighted recommendation. |
meeting-follow-up-pipeline/SKILL.md |
Turns a finished meeting into a debrief, decisions, action items, ready-to-send follow-up drafts, and proposed writebacks into your task/CRM systems. |
exec-weekly-agenda-generator/SKILL.md |
Sweeps your week across meetings, chat, email, and calendar, then produces a forced-ranked list of 5-10 topics worth the leadership team's time. |
In Claude Code: install as a plugin — claude plugin marketplace add zapier/marketplace then claude plugin install wade-skills@zapier. All three skills become available automatically.
In OpenAI Codex: codex plugin marketplace add zapier/marketplace then codex plugin add wade-skills@zapier. Or open the in-CLI picker with /plugins and toggle it on.
In GitHub Copilot CLI: copilot plugin marketplace add zapier/marketplace then copilot plugin install wade-skills@zapier.
In Cursor or Claude Code (manual): drop a skill's folder into .cursor/skills/ (Cursor) or .claude/skills/ (Claude Code). The YAML frontmatter at the top of each file (name + description) is what tells the agent when to invoke it automatically. You can also just paste the file contents into a chat and say "follow this."
In Claude Projects / a custom GPT / any chatbot: paste the skill file into the system prompt or project knowledge. Then trigger it with the phrases listed under "When to Use."
Fill in the placeholders first. Each file has a short "Setup" section listing the [BRACKETED] placeholders (your meeting-notes tool, your email, your CRM, your leadership team roster, etc.). Set those once and the skill runs against your stack.
These work because of a few principles, not because of any specific vendor:
- Personas with flaws give better critiques than generic advisors. The War Council members have backstories that include failures on purpose.
- Every output must trace to a real source. The agenda generator refuses to include a topic it can't tie to an actual meeting, thread, or email. No generic "review metrics" filler.
- Propose, don't auto-execute. The follow-up pipeline drafts and proposes writebacks but asks before sending or writing anything.
- Force ranking beats completeness. Five strong items beat ten mediocre ones.
Shared with the Running Remote community and released by Zapier, Inc. under the MIT License. See LICENSE for the license and NOTICE for trademark, third-party, and use terms. Review NOTICE before use.
This repo is shared as-is. We're not accepting external contributions.