Advisory Committee

4-agent AI advisory committee for strategic decisions. Runs CS, Marketing, Finance, and Tech agents in sequence to analyze proposals and deliver a majority-v...

Security Scan

Instruction-only skill that simulates a four-member advisory committee with no installs, credentials, or external endpoints — its requirements and instructions are coherent with its stated purpose.

Purpose & Capability

Name/description (4-agent advisory committee) matches the SKILL.md and the included role definitions; no unexpected binaries, env vars, or external services are requested.

Instruction Scope

Runtime instructions stay within the advertised purpose (run four agents in sequence, produce per-agent verdicts, and append a decision-log row). Minor ambiguity: the skill doesn't specify tie-breaking rules, how much context each agent may use, or where the decision log is persisted (conversation only, memory store, or external storage). This is a functional omission rather than a safety mismatch.

Install Mechanism

No install spec and no code files — lowest-risk model. Nothing is downloaded or written to disk by the skill itself.

Credentials

No environment variables, credentials, or config paths are requested; requested permissions are proportionate to the described functionality.

Persistence & Privilege

always:false and autonomous invocation allowed (platform default). The skill does not request elevated or persistent platform privileges or modify other skills. Consideration: outputs/logs could be persisted by the host agent if that host saves conversation history or memory.

Assessment

This skill is instruction-only and appears internally consistent with its stated purpose. Before enabling it broadly, confirm where the decision log will be stored (conversation only vs. persistent memory or external storage), how tie-breaks are resolved, and whether your agent will run the committee autonomously on triggers you care about. Avoid submitting sensitive or confidential proposals until you've verified the host's persistence and sharing policies, and test with non-sensitive examples first.

How It Works

When given a proposal, run the following 4 agents in sequence. Each agent judges independently.

Agent Order

1. 🎧 CS Agent "User Champion" — User value perspective
2. 📣 Marketing Agent "Growth Hacker" — Distribution/growth perspective
3. 📦 Finance Agent "CFO" — ROI/priority perspective
4. 🔧 Tech Agent "CTO" — Implementation/complexity perspective

Each agent verdict: ✅ YES / ❌ NO / ⚠️ Conditional

Output Format

## 🏛️ Committee — [Proposal]

### 🎧 CS — [✅/❌/⚠️]
> [2~3 lines of reasoning]

### 📣 Marketing — [✅/❌/⚠️]
> [2~3 lines of reasoning]

### 📦 Finance — [✅/❌/⚠️]
> [2~3 lines of reasoning]

### 🔧 Tech — [✅/❌/⚠️]
> [2~3 lines of reasoning]

## 📋 Recommendation
**Decision: [YES / NO / Conditional]**
> [One-line action]

Agent Criteria

For detailed judgment criteria per agent, see references/committee-roles.md.

Decision Log

After each decision, append a row to the decision log table:

Date	Proposal	CS	Marketing	Finance	Tech	Decision