acmm-auditlisted

# ACMM Audit Canonical AI Codebase Maturity Model — scores how AI-operable a repo is on a 6-level scale (L1 Assisted → L6 Fully Autonomous). Fills the gap between `/site-audit` (UX/code health) and `/progress-tracker` (loop metrics): ACMM evaluates the **meta-properties** of the repo — do we have the instructions, metrics, loops, and gates in place to be AI-driven? The criterion catalog is **ported verbatim** from [kubestellar/console](https://github.com/kubestellar/console/tree/main/web/src/lib/acmm/sources) — the reference implementation validated in the [arXiv paper (2604.09388)](https://arxiv.org/abs/2604.09388). 100+ criteria total across 4 cited source frameworks. ## Invocation ```bash # Dry run — scores the repo, writes report, creates nothing node ${CLAUDE_PLUGIN_ROOT}/scripts/audit.js # Create deduplicated GitHub issues for next-level gaps (fed to ship-loop) node ${CLAUDE_PLUGIN_ROOT}/scripts/audit.js --apply # Rewrite README badge between <!-- acmm:begin -->/<!-- acmm:end --> node ${CLAUDE_PLUGIN_ROOT}/scripts/audit.js --badge # Full run (scheduled trigger invokes this) node ${CLAUDE_PLUGIN_ROOT}/scripts/audit.js --apply --badge # Just print trend from .claude/acmm/state.json node ${CLAUDE_PLUGIN_ROOT}/scripts/audit.js --trend ``` ## What it does 1. Loads 100+ criteria from `${CLAUDE_PLUGIN_ROOT}/scripts/sources/{acmm,fullsend,agentic-engineering-framework,claude-reflect}.js`. 2. Runs file-presence detection on each (no network, native `fs` only): - `pa