mneves75

agent-readiness

1
0
# Install this skill:
npx skills add mneves75/skills --skill "agent-readiness"

Install specific skill from multi-skill repository

# Description

Evaluate codebase readiness for AI coding agents using automated assessment. Use when onboarding repos, diagnosing agent struggles, or planning infrastructure improvements.

# SKILL.md


name: agent-readiness
description: Evaluate codebase readiness for AI coding agents using automated assessment. Use when onboarding repos, diagnosing agent struggles, or planning infrastructure improvements.


Agent Readiness Evaluation

When to Use This Skill

  • Onboarding a new repository for AI-assisted development
  • Diagnosing why your AI agent is struggling with a codebase
  • Evaluating a codebase before AI agent deployment
  • User asks "why isn't my AI working well here?"
  • User asks about "agent readiness" or "codebase evaluation"
  • Planning infrastructure improvements for agent performance

Core Insight

"The agent is not broken. The environment is."

AI coding agents operate in a feedback loop: gather context -> take action -> verify work -> iterate

This skill evaluates whether your codebase supports each phase of that loop.

Quick Assessment (5 Questions)

Answer these for an instant maturity estimate:

  1. Does CLAUDE.md (or .cursorrules, AGENTS.md) exist with build/test commands? (CONTEXT)
  2. Is strict typing enabled with linting? (ACTION)
  3. Can tests run without manual setup? (VERIFICATION)
  4. Does CI provide clear pass/fail feedback? (ITERATION)
  5. Is the feedback loop under 5 minutes? (AUTONOMOUS)

Score: 5/5 = L5, 4/5 = L4, 3/5 = L3, 2/5 = L2, 1/5 = L1

Automated Assessment

Run the readiness-check tool (v1.0.0):

# From the skills repository
cd ~/.claude/skills/mneves-skills
bun --bun tools/readiness-check.ts /path/to/your/project

# Quick mode (skip tests/build)
bun --bun tools/readiness-check.ts --skip-tests --skip-build

# HTML dashboard
bun --bun tools/readiness-check.ts --format=html --output=readiness-report.html

# CI gate (fail if below level 3)
bun --bun tools/readiness-check.ts --min-level=3

Four Phases to Evaluate

Phase Question Key Checks
CONTEXT Can the agent understand? CLAUDE.md exists, <300 lines, has commands
ACTION Can the agent change things? No any, strict types, linting configured
VERIFICATION Can the agent check work? Tests pass, build works, lockfile committed
ITERATION Can the agent improve? CI configured, PR template, required checks

Maturity Levels

Level Name Description
L1 Functional Code runs, basic files exist
L2 Navigable Agent can find things
L3 Actionable Agent can make changes safely
L4 Verifiable Agent can check its work
L5 Autonomous Agent can work independently

Target: Level 3 minimum for productive AI development.

Remediation Paths

Gap Action
Missing context file Create CLAUDE.md/.cursorrules with build/test commands
No typing/linting Add ESLint/Biome + strict TypeScript
Tests require setup Use containers or in-memory alternatives
No CI pipeline Add GitHub Actions workflow
No observability Add structured logging

Context File Template

Works as CLAUDE.md, .cursorrules, or AGENTS.md:

# Project Context

## Overview
[1-2 sentences: what this does and WHY]

## Commands
- Build: `bun run build`
- Test: `bun test`
- Lint: `bun run lint`

## Architecture
[Brief description of key directories]

## Key Decisions
- [Decision]: [Why]

Requirements: Concise (<300 lines), Actionable (commands work), Current (updated within 30 days)

Compliance Checklist

CONTEXT Phase

  • [ ] Context file exists at repository root (CLAUDE.md, .cursorrules, AGENTS.md)
  • [ ] Context file includes build/test commands
  • [ ] Context file is under 300 lines
  • [ ] File structure is recognizable (src/, app/, lib/)

ACTION Phase

  • [ ] No any type casts in codebase (if TypeScript)
  • [ ] Strict mode enabled
  • [ ] Linter configured (ESLint, Biome, golangci-lint, ruff)
  • [ ] Formatter configured

VERIFICATION Phase

  • [ ] Tests exist and pass
  • [ ] Build succeeds without manual intervention
  • [ ] Lock file is committed

ITERATION Phase

  • [ ] CI pipeline configured
  • [ ] PR template exists
  • [ ] Deployment process documented

Common Mistakes

  • Blaming the agent when infrastructure is broken (fix the environment!)
  • Missing context file (single most impactful file for agent success)
  • Liberal use of any to silence TypeScript (use unknown + type guards)
  • Tests that require manual setup (use containers or in-memory DBs)
  • Outdated documentation (keep context files current)

Reference

"You can't prompt your way out of bad infrastructure. Fix the environment, not the agent."

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.