flow-next-prime_skill

Overview

This skill performs a comprehensive codebase assessment focused on agent and production readiness. It scans eight pillars (48 criteria), verifies that commands actually run, inspects GitHub settings, and reports findings. It offers targeted fixes only for agent readiness pillars to make repositories reliably usable by autonomous agents.

How this skill works

The skill runs nine lightweight scouts in parallel to collect fast diagnostics across style, build, testing, docs, dev environment, observability, security, and workflows. It then verifies executable commands (tests, build, linters), computes an agent-readiness score and maturity level, and produces a full report. For Agent Readiness pillars (1–5) it proposes and can generate config/docs/scripts after explicit user consent; pillars 6–8 are reported only.

When to use it

• Preparing a repo so AI agents can work reliably and autonomously
• Evaluating a project before enabling overnight/autonomous coding (Ralph mode)
• Prioritizing immediate fixes that reduce agent failures and wasted cycles
• Verifying that test/build/lint commands actually run in the repository environment
• Generating an actionable agent-focused remediation plan before scaling automation

Best practices

• Run the assessment from the repository root or pass an explicit path to scan a different project
• Use --report-only to gather data without applying fixes, useful for audits
• Prefer incremental fixes: aim for Maturity Level 3 (Standardized) before optimizing further
• Always review generated config/docs/scripts before committing; the skill never commits changes for you
• Provide explicit consent when asked; the tool uses an AskUserQuestion step before any modifications

Example use cases

• Onboard a new autonomous agent pipeline and ensure fast local feedback (pre-commit, linters, type checks)
• Hardening developer experience: add .env.example, devcontainer, and test commands so agents can reproduce runs
• Audit a repo for missing CLAUDE.md or agent guidance and generate a concise CLAUDE.md draft
• Run nightly checks to detect drift in test/build commands and re-anchor repository context
• Prepare a monorepo for agent workflows by verifying build tooling and lockfiles