Guide · Agent skill
How to use the Chief of Staff Orchestrator agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Chief of Staff skill (skill name chief-of-staff) turns an AI agent into a coordinator for a startup founder's executive questions. It decides which simulated C-suite role should answer, convenes a multi-role board meeting when a decision is complex, and combines the answers into one summary. This guide explains what the skill does, how to install it, how to trigger it, and where its limits are.
Chief of Staff Orchestratoralirezarezvani/claude-skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the chief-of-staff ZIP from this page.
- 2Unzip the folder. Unzip the file so you have a chief-of-staff folder containing SKILL.md and references/.
- 3Place the skill. Move the folder into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 4Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 5Ask a business question. Describe your decision in plain language, such as whether to raise now or cut burn.
- 6Review the synthesis. Read the agreements, disagreements, actions and the single decision point, then make the call yourself.
What the Chief of Staff skill makes the agent do
The skill is an orchestration layer. When you ask a business question, the agent first loads company context through the context-engine skill, scores how complex the decision is, and then either routes the question to one or two roles or triggers a board meeting. Roles include CFO, CTO, CMO, CHRO, General Counsel and others, as listed in its routing matrix.
Routing is based on a table of topics. For example, fundraising, burn and financial models go to the CFO with the CEO as secondary, while security and compliance go to the CISO with the COO as secondary. Complexity scoring decides the depth: a single-domain question gets one role, a question where two domains intersect gets two roles, and a question spanning three or more domains with major trade-offs or irreversible effects gets a board meeting.
- Single domain, clear answer: one role answers.
- Two domains intersect: two roles answer and the output is synthesized.
- Three or more domains, major trade-offs or irreversible choices: a board meeting is convened.
- Each role gets one turn in a board meeting; roles do not debate each other directly.
Install the chief-of-staff skill from the ZIP
Download the ZIP for this skill from this page, then unzip it. Put the resulting chief-of-staff folder into your personal skills directory, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill only for that project. The folder should contain SKILL.md and its references/ directory.
If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/chief-of-staff/
- Project install: .claude/skills/chief-of-staff/ inside your project
- Claude apps: Settings → Capabilities → Skills
~/.claude/skills/chief-of-staff/Trigger the skill with natural requests
You do not need special syntax to start. The skill's description says to use it when a founder question needs routing to the right advisor, or when a multi-domain decision needs a board meeting convened. Ask the question in plain language, such as 'should we raise now or cut burn?'.
The skill file also documents an internal invocation format, [INVOKE:role|question], which the orchestrator uses to call roles. You can read it to understand the output, but you do not need to type it yourself.
- 'Should we raise now or cut burn?' (routes to CFO and CEO)
- 'Should we hire or extend runway?' (routes to CFO, with CHRO and COO)
- 'Convene a board meeting on whether to pivot our pricing model.' (a board meeting for a multi-domain decision)
[INVOKE:cfo|What's the right runway target given our growth rate?]Read the output: agreements, disagreements and decision points
For multi-role answers, the skill is designed to produce a fixed structure: What We Agree On, The Disagreement, Recommended Actions, and Your Decision Point. Conflicts are named rather than smoothed over, and the decision point is framed as two options with trade-offs and no recommendation, so the final choice stays with you.
Actions are meant to have an owner and a timeline. The skill also keeps a decision log: each decision is recorded in ~/.claude/decisions/, and at session start the agent checks approved decisions and flags any whose review date has passed.
- What We Agree On: two to three themes that multiple roles reached independently.
- The Disagreement: the named conflict, each side's reasoning, and what it is really about.
- Recommended Actions: up to five actions, each with an owner and timeline.
- Your Decision Point: one question with two options.
## Your Decision Point
[One question. Two options with trade-offs. No recommendation — just clarity.]Requirements and safety notes
The skill needs an agent that supports SKILL.md skills, such as Claude Code or the Claude apps. It also needs write access to ~/.claude/decisions/ for the decision log. The companion skills it references (context-engine, agent-protocol, decision-logger, board-meeting and the other C-suite role skills) are not included in these files, so the orchestrator is not complete on its own.
The skill does not run scripts and does not use the network, according to this listing. Its own files do not say where company context comes from, so answers may be generic if the context is missing. The output is AI-generated advice from simulated executive roles and is not reviewed by qualified professionals.
- Runs scripts: no
- Uses network: no
- Needs write access to ~/.claude/decisions/
- Companion skills must be installed separately for full functionality
Tips for better results
Give the agent concrete facts in your question, such as runway in months, current burn and the specific choice in front of you. The skill is designed to load company context, but the files do not define its source, so check that the answer uses your real numbers.
Use the board meeting for decisions that touch several functions or cannot be easily reversed. For a simple factual question, a single role answer is faster and easier to review.
- Name the decision and the options you are weighing.
- Include the constraints that matter, such as deadlines or team size.
- Save important decisions so the review-date check can flag them later.
- Treat outputs as input to your own judgement, not as a final answer on legal, financial or compliance questions.
What people use it for
- Deciding whether to raise a round now or extend runway by cutting costs, with the CFO and CEO perspectives combined.
- Weighing a pricing model change across finance, revenue and product roles.
- Making a hire-or-extend-runway decision with input from people and operations roles.
- Testing how a multi-agent advisory workflow built on SKILL.md files behaves, as a developer.
- Logging founder decisions with owners and review dates in ~/.claude/decisions/.
Frequently asked questions
How do I install the chief of staff agent skill?
Download the ZIP from this page and unzip it. Place the chief-of-staff folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings → Capabilities → Skills.
Does the chief-of-staff skill need anything else installed?
Yes. It references companion skills that are not included in these files: context-engine, agent-protocol, decision-logger, board-meeting and the other C-suite role skills. Without them the orchestrator is incomplete.
Is the chief-of-staff skill free to use?
The skill is listed under the MIT licence. This page does not state any pricing, so check the licence file and your agent platform's terms for any costs.
Does the skill run scripts or access the internet?
According to this listing, it runs no scripts and uses no network. It does need write access to ~/.claude/decisions/ to keep the decision log.
Can I use this instead of a human lawyer or financial advisor?
No. The output is generated advice from an AI playing executive roles and is not reviewed by qualified professionals. It can sound confident while being wrong, so use it to structure your thinking and get expert review for legal, financial or compliance decisions.
How is this different from asking a single AI assistant a business question?
The skill routes each question to specific role perspectives, names disagreements between them, and frames the final choice as options with trade-offs instead of one recommendation. It also keeps a decision log and flags review dates.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits founders or small leadership teams who want structured trade-off analysis across finance, product, tech, hiring and legal, and developers studying multi-agent SKILL.md workflows. It is a poor fit if you need verified company data or a standalone product, since it depends on companion skills not included here and its context source is undefined. For legal, financial or compliance decisions, use it only as a framing aid and rely on qualified professionals.
Download the skill