Guide · Agent skill
How to use the Inter-Agent Protocol agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Inter-Agent Protocol skill (agent-protocol) is a rulebook for simulated executive roles, such as CEO, CFO and CTO, when they need to ask each other questions. It defines a standard request syntax, a standard reply format, and rules that stop agents from calling each other in endless loops. This guide explains what the skill changes, how to install it, how to trigger it, and where its limits are.
Inter-Agent Protocolalirezarezvani/claude-skills · MIT Download the skill Quick start
- 1Download the ZIP. Get the agent-protocol ZIP from this page and unzip it.
- 2Choose a location. Place the folder in ~/.claude/skills/ for all projects, or in .claude/skills/ inside one project.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Ask a cross-functional question. Describe a task involving several executive roles, such as budget and hiring together.
- 5Check the chain and assumptions. Review each [RESPONSE] block, its confidence and caveat, and any [ASSUMPTION] or [CONFLICT] flags.
- 6Review before acting. Have a human verify the numbers before using any output for a real decision.
What the agent-protocol skill makes the agent do
Agent skills are folders of instructions that an AI assistant loads when a task matches them. This one covers communication between simulated C-suite roles. Without it, roles have no agreed way to ask each other for input, reply, or stop a chain of questions from running forever.
With the skill loaded, invocations use the form [INVOKE:role|question], replies use a [RESPONSE:role] block with a key finding, supporting data, a confidence level and a caveat, and every hop carries a call chain so agents can check for loops.
- Valid role tokens include ceo, cfo, cro, cmo, cpo, cto, chro, coo, ciso, gc, cdo, caio, cco and vpe.
- Hard loop rules: no self-invocation, maximum chain depth of 2, and no circular calls.
- Isolation rules apply during board meeting phases, and conflicts between roles must be surfaced rather than silently resolved.
Install the agent-protocol skill
Download the ZIP for this skill from this page and unzip it. You can install it in one of two ways, depending on where you work.
For Claude Code, place the skill folder in ~/.claude/skills/ to make it available everywhere, or in a project's .claude/skills/ folder to limit it to that project. The skill folder should contain SKILL.md and references/invocation-patterns.md.
In the Claude apps, upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/agent-protocol/
- Project install: .claude/skills/agent-protocol/ inside your project
How to trigger the skill with natural requests
You do not need a special command. The skill's description says it applies when C-suite agents need to query each other, coordinate cross-functional analysis, or run board meetings with multiple agent roles. Ask for that kind of work in plain language and the assistant can load the skill.
Requests that fit the skill involve more than one role, such as asking how a hiring plan affects both budget and delivery capacity, or asking a CEO-style agent to run a board meeting with CFO and CRO input.
- Ask a cross-functional question: "Have the CFO and CRO cross-check whether our Q3 hiring plan is affordable given the pipeline."
- Ask for a board-style review: "Run a board meeting on the fundraise timing with independent analysis first, then a critique."
- Ask for a broadcast: "Ask every executive role what happens if we miss the fundraise."
[INVOKE:cfo|What's the burn rate impact of hiring 5 engineers in Q3?]Read a response and handle conflicts
Each reply follows a fixed structure, so you can scan it quickly: the key finding comes first, then the supporting data, then the confidence level and the caveat that could make the answer wrong.
When two roles disagree, the protocol says to flag the conflict explicitly, state how it will be resolved (conservative, probabilistic, or escalate to a human), and never silently pick one. Treat any stated assumption, marked [ASSUMPTION: ...], as something to check.
- Confidence values are high, medium or low.
- Blocked calls return a [BLOCKED: ...] message with the assumption used instead.
- Conflicts appear as [CONFLICT: ...] lines.
[RESPONSE:cfo]
Key finding: [one line — the actual answer]
Supporting data:
- [data point 1]
Confidence: [high | medium | low]
Caveat: [one line — what could make this wrong]
[/RESPONSE]Understand the safety and data limits
The skill defines message rules only. It does not connect to live company data. The source states it assumes other skills supply the content, so the numbers in a response may be assumptions rather than checked facts.
The skill's listed requirements are nothing, it runs no scripts, and it uses no network access. Its loop and isolation rules are enforced by the instructions the assistant follows, not by a separate program, so treat them as guidance for the assistant rather than a technical guarantee.
- Requires: nothing listed.
- Runs scripts: no.
- Uses network: no.
- Do not use outputs for real business decisions without human review, because the roles are AI simulations.
Examples of common patterns
The reference file references/invocation-patterns.md contains worked examples. The revenue planning chain runs CRO → CFO → CMO: the CRO asks the CFO for a Q3 revenue target given runway and burn, then asks the CMO what pipeline coverage and acquisition cost that target implies.
Each step in a chain is logged with its call chain, for example [CHAIN: cro → cfo → coo], so the depth limit can be checked at each hop.
What people use it for
- Testing how a hiring plan affects both runway and delivery capacity before a planning meeting.
- Running a structured board-style review in which roles analyse independently before critiquing each other.
- Studying how loop prevention and isolation rules can be designed for multi-agent workflows.
- Building a multi-agent executive advisory setup in Claude Code or a similar tool.
Frequently asked questions
Is the agent-protocol skill free?
The skill is released under the MIT licence, as stated on this page. Check the licence file in the repository for the full terms before reusing it.
What does the agent-protocol skill require?
The source lists no requirements. It runs no scripts and uses no network access. It does need the other C-suite advisor skills to supply domain content, since it only defines message rules.
How do I install the agent-protocol skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings → Capabilities → Skills.
Why is my agent blocked from calling another role?
The protocol blocks self-invocation, chains deeper than two hops, and circular calls. It also blocks invocations during board meeting Phase 2 and during Phase 3 critique. A blocked call returns a BLOCKED message and the agent should state the assumption it used.
Can the skill pull live company data?
No. The source says the skill only defines message rules and assumes other skills supply the content. Numbers in responses may be unverified, so check them yourself.
How does this compare with using a single assistant?
For a simple one-question chat with a single assistant, the source says the protocol adds overhead without benefit. It is useful when you want several advisor roles to cross-check each other's analysis.
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 developers building multi-agent executive advisory setups and teams studying loop prevention and isolation design. It is not a good fit for simple single-assistant questions, for decisions made without human review, or for anyone expecting live company data to be connected automatically.
Download the skill