Inter-Agent Protocol

Defines the inter-agent communication protocol for C-suite agent teams, including invocation syntax, loop prevention rules, isolation rules, response formats, and verification standards. It ensures structured, non-circular coordination among role agents and produces standardized outputs for the founder.

✨ What it does

  • Defines invocation syntax [INVOKE:role|question] and valid role tokens.
  • Enforces loop prevention rules: no self-invocation, max depth 2, no circular calls, chain tracking.
  • Specifies isolation rules for board meeting phases (no invocations in Phase 2, reference-only in Phase 3).
  • Provides response format [RESPONSE:role] with key finding, supporting data, confidence, and caveat.
  • Outlines conflict resolution and broadcast patterns for CEO.
  • Establishes a decision memory layout and internal quality loop with self, peer, and critic verification.

🎯 When to use it

  • When C-suite agents need to query each other for domain-specific data or cross-functional analysis.
  • When running board meetings with multiple agent roles that require independent analysis and critique phases.
  • When coordinating multi-agent workflows that need to prevent loops, circular calls, or excessive chain depth.
  • When standardizing the format of outputs or alerts delivered to the founder.

🚀 How to use

Trigger: /cs:decide

Trigger by using the invocation syntax within agent interactions, e.g., when a C-suite agent needs data from another role. The skill expects a question directed at a valid role token. Example prompts:

[INVOKE:cfo|What's the burn rate impact of hiring 5 engineers in Q3?]
[INVOKE:cto|Can we realistically ship this feature by end of quarter?]

For board meetings, follow the phase-specific isolation rules. The skill also applies to any output to the founder, which must follow the standard communication format.

📄 Output: Standardized inter-agent messages, verified outputs, and founder-facing reports following the defined formats.

📦 Add this skill to Claude Code

# 1. Get the skills repo
git clone --depth 1 https://github.com/alirezarezvani/claude-skills /tmp/claude-skills

# 2. Copy this skill into your project (or ~/.claude/skills for all projects)
mkdir -p .claude/skills
cp -r /tmp/claude-skills/c-level-advisor/skills/agent-protocol .claude/skills/agent-protocol

Skill source: c-level-advisor/skills/agent-protocol/SKILL.md

⚠️ Good to know

The protocol is designed for C-suite agent teams and assumes adherence to the defined roles and rules; it does not cover non-C-suite agents or scenarios outside the specified phases.

❓ FAQ

What happens if an agent tries to invoke itself?

Self-invocation is always blocked; the agent must state an assumption instead.

Can an agent invoke another agent during board meeting Phase 2?

No, Phase 2 isolation prohibits all invocations; agents must state assumptions with [ASSUMPTION: ...].

How are conflicting answers from two agents resolved?

Conflicts are flagged explicitly, and the resolution approach is stated (conservative, probabilistic, or escalate); never silently pick one.

🤖 Overview, features, install steps and FAQ were generated from the project's SKILL.md on Sep 4, 2026. Always check the original source before running commands.