Agentfit

Agent-Fit: Agent-Readiness Assessment Skill

A Claude Code skill that evaluates your codebase's agent-readiness across 9 technical pillars, 86+ binary criteria, and 5 maturity levels. Produces an HTML report with pass rate scoring and a prioritized remediation roadmap.

Install

Add to your Claude Code settings (.claude/settings.json or global):

{
  "extraKnownMarketplaces": {
    "agent-fit": {
      "source": {
        "source": "git",
        "url": "git@github.com:HarshwardhanPatil07/agent-fit.git"
      }
    }
  },
  "enabledPlugins": {
    "agentfit@agent-fit": true
  }
}

Run

Inside Claude Code, in any project directory:

/agentfit

What You Get

An HTML report and matching JSON sidecar scoring your project 0-100% across 9 pillars:

Pillar	Criteria	What It Measures
Style & Validation	13	Formatting, linting, type checking, complexity
Build System	19	CI/CD, dependencies, releases, automation
Testing	8	Coverage, isolation, flaky detection, performance
Documentation	10	README, AGENTS.md, API docs, architecture, changelog
Dev Environment	5	Devcontainer, env templates, local services
Debugging & Observability	11	Logging, tracing, metrics, alerting, profiling
Security & Governance	13	Branch protection, secrets, scanning, CODEOWNERS, SBOM
Task Discovery	4	Issue templates, labeling, backlog health
Product & Analytics	3	Analytics, error pipelines, A/B testing

Output Files

• HTML: /tmp/agentfit-report-{project_name}.html
• JSON: /tmp/agentfit-report-{project_name}.json

The JSON sidecar includes report metadata, pillar and criterion details, level progress, gate status, and delta fields for trend automation.

Report UX Features

• Collapsible pillar sections
• Status filters: Show All, Found, Missing, Skipped
• Criterion name search
• Level filter (L1-L5)
• Metadata footer (timestamp, version, criteria totals, git SHA, duration)
• Delta highlighting for criteria that changed since the previous JSON baseline

Maturity Levels

Level	Name	What It Means
L1	Functional	Code runs but requires manual setup
L2	Documented	Basic docs and processes exist
L3	Standardized	Processes enforced through automation
L4	Optimized	Fast feedback loops, data-driven improvement
L5	Autonomous	Self-improving systems

To unlock a level, pass 80% of that level's criteria AND all previous levels.

Console output also includes per-level gate status (✓, ✗, — blocked) so CI logs show why a level is locked.

Supported Languages

Go, Python, TypeScript/JavaScript, Rust, C++, Swift

Optional: GitHub API Criteria

Some criteria (branch protection, issue labels, secret scanning) use the gh CLI. If gh is not installed or authenticated, those criteria are gracefully skipped.

Trend Comparison

When /tmp/agentfit-report-{project_name}.json exists from a previous run, Agent-Fit computes pass-rate delta and highlights criterion status changes. If no baseline is available (or baseline is malformed), the run still succeeds and reports current-state results.

Requirements

• Claude Code (that's it — no dependencies, no API keys, no installs)