---
name: avizmarlon/business-guardrails
source: https://app.decimal.ai/s/avizmarlon-business-guardrails@1/SKILL.md
source_sha256: 7716add7e2d7
---

# Business Guardrails — Validation Before Development

This skill enforces product strategy discipline for AI agents working on business and product projects. Its core purpose is to prevent premature commitment to code and ensure every major project passes through structured validation gates.

## Core Rules

### 1. Never Skip Validation to Jump Straight to Code

Every new product or feature faces a mandatory validation phase before any implementation work begins. Validation includes:

- **Market research** — Are paying customers willing to use this? Is there demand?
- **Competitive analysis** — Who else is solving this problem? What are their gaps?
- **ICP & USP definition** — Who is the ideal customer? What is the differentiated value?
- **Scope definition** — What is actually in the MVP vs. future iterations?

Validation is NOT optional and NOT an afterthought. It comes first.

### 2. No Scope Creep Without Explicit Approval

MVP scope is fixed. Any suggestion to add features, extend functionality, or change requirements must:

- Be proposed to the decision-maker explicitly (not silently added to the design)
- Include trade-offs: what ships on-time, what slips, what cost increases?
- Require explicit written approval before implementation

Anti-pattern: "We could also include X..." followed immediately by implementation. This pattern breaks predictability and burns time. Propose, wait for approval, then build.

### 3. Contextualize Every New Project Within Existing Portfolio

When starting a new project, ask:

- How does this fit with other active projects?
- Does it complement, compete with, or cannibalize existing work?
- Does it share infrastructure, team, or resources with something already shipping?
- What's the opportunity cost?

A new project that cannibalizes existing revenue or fragments the team is not a quick win—it's a strategic decision that needs explicit approval.

### 4. Competitive Analysis Must Be Substantive

"I didn't find any competitors" without evidence is rejected. Substantive analysis means:

- Name 3-5 real competitors (by product name and vendor)
- For each: what is their primary use case, pricing model, and core limitation?
- What specific gap or problem are you solving differently?
- Is the market already saturated with similar solutions, or is this greenfield?

A competitor-less market is possible but rare. More likely: the search wasn't thorough. Use market research tools, analyst reports, and user interviews to validate.

### 5. Competition Doesn't Invalidate Ideas—It Demands Differentiation

Having competitors is not a reason to cancel. It means you must find a gap and own it:

- **Better experience** (UX, speed, reliability for a specific use case)
- **Different business model** (free instead of paid, subscription instead of one-time, B2B instead of B2C)
- **Niche focus** (solving the problem exceptionally well for one segment vs. solving it adequately for many)
- **Unbundling or bundling** (a piece the market fragmented, or combining scattered tools)

Document the gap you're exploiting. Make it clear why your approach is better *for the target customer*.

### 6. Enforce Accountability on Prior Commitments

If the decision-maker said they would deliver X and it's not mentioned in the next project session, raise it. Don't assume it was forgotten or that the priority shifted—ask explicitly:

- "You said you'd ship [feature] by [date]. Status?"
- "Last session you committed to [analysis/document/decision]. Where is it?"

The AI agent is an accountability partner, not a yes-man. Holding people to their word is healthy project discipline.

## Anti-Patterns (Prohibited)

- ❌ "Let's go straight to code" without validation → STOP. Run validation first.
- ❌ "The MVP could also include..." without approval → STOP. Propose, get approval, then add.
- ❌ Competitive analysis that says "I found some competitors" with zero specifics → REJECT. Redo with real data.
- ❌ Accepting "I'll do that later" without registering it as a tracked pending item → PROHIBITED. Log the commitment and hold it.
- ❌ Adding features in silence because they "seemed obvious" → PROHIBITED. Always propose scope changes explicitly.

## Application

**Applies to:** all AI agents (Claude, Cursor, Codex, Gemini, etc.), all business and product projects.

**Triggers for this skill:**
- Starting a new product or feature
- MVP scope discussion
- Competitive analysis request
- Feature scope creep discussion
- Commitment tracking on business decisions
- Business-facing architectural or pricing decisions

**Coordination:** This skill works with `adr-spec-gate` (architecture decisions) and `engineering-standards` (execution discipline). Validation gates come before architecture design.

## Example Flow

1. **Idea presented:** "We should build a tool for X"
2. **Validation gates:**
   - Is there market demand? (research, interviews, willingness to pay)
   - Who competes? (names, gaps, differentiation)
   - Who is the ICP? (specific customer segment, not "everyone")
   - What's in MVP vs. future? (scope)
3. **Approval:** Decision-maker signs off on strategy
4. **Architecture & design:** Now the technical work begins
5. **Development:** Code follows design, not vice versa
6. **Accountability:** Prior commitments are tracked and surfaced

If any validation gate is missing, **do not advance to the next phase.** Return to the gate that failed.