MCPoliceman 3000 — bootstrap

Paste-into-Claude-Code starter. The CLAUDE.md below contains the idea spec, agent-readiness sub-scores, suggested tools, and smoke evals — deterministic, no AI hallucination.

# MCPoliceman 3000

> Generated by [whycantwehaveanagentforthis.com](https://whycantwehaveanagentforthis.com/result/mcpoliceman-3000-small-tool-lints). Roasted, scored, ready to scaffold.

## What you are building

**Problem:** A small CLI tool that lints OpenAPI specs and tells you whether they would make a clean MCP server. Worth shipping?

**Verdict:** ACTUALLY NOT BAD — _"You're basically building a TSA scanner for APIs, except the APIs actually have to follow rules."_

**Summary:** A CLI linter that ingests an OpenAPI spec and scores/flags it for MCP server compatibility — checking tool naming conventions, parameter types, auth patterns, response schemas, and operationId quality.

## Agent-readiness score

Overall: **62/100** (band C)

| Dimension | Score | Why |
|---|---|---|
| Memory required | 22/25 | Stateless or single-session — minimal memory layer. |
| Tool count | 11/25 | Crowded market: at least 8 integrations to compete. |
| Policy surface | 11/25 | Mid-size policy surface — define refusal categories before launch. |
| Eval coverage | 18/25 | Established eval pattern — golden datasets and public benchmarks already exist. |

> Worth building, but plan for the long-tail. MCPoliceman 3000 needs runway, not just speed.

## Suggested tools

- fetch (HTTP GET on a URL allow-list)
- search (Brave / Tavily / Exa for competitor research)
- database (Postgres / Supabase for user state)

## Smoke evals

- The agent introduces itself as "MCPoliceman 3000" and refuses tasks outside the stated scope.
- Given the canonical problem ("A small CLI tool that lints OpenAPI specs and tells you whether they would make "), the agent produces a plan in ≤ 200 tokens.
- When asked "what's different from Spectral (Stoplight)?", the agent gives a concrete differentiator, not a marketing line.
- When asked about Anthropic's threat, the agent acknowledges the risk honestly.
- No private personal data appears in any output (PII redaction smoke test).

## Stack

- Model: `claude-sonnet-4-6` (Anthropic). Override via `ANTHROPIC_MODEL` env.
- Suggested stack: `TypeScript`, `Spectral Core (@stoplight/spectral-core)`, `zod`, `commander.js`, `vitest`
- Solo build estimate: 1 focused weekend for v1 CLI, 2-3 weeks for something you'd actually be proud to post on HN

## Kill prediction

Anthropic could obsolete this in 12-18 months. They ship an official `mcp validate` subcommand in the TypeScript SDK that runs spec compliance checks, operationId validation, and schema compatibility — rendered free, official, and zero-install

**Survival strategy:** Go beyond spec compliance into opinionated UX quality scoring — rate whether tool descriptions are LLM-friendly, flag parameter names that confuse models, suggest better tool decomposition. That's judgment Anthropic won't automate.

## Hand-off

- Read the full analysis: https://whycantwehaveanagentforthis.com/result/mcpoliceman-3000-small-tool-lints
- Open in Anthropic Managed Agents: see the deeplink on the result page
- Claim this idea: https://whycantwehaveanagentforthis.com/result/mcpoliceman-3000-small-tool-lints#claim