Repository Agent-Readiness Audit

Perform a read-only audit of a code repository to determine how well it supports autonomous AI coding agents. Produce an Agent-Readiness Report covering 86 checks grouped into 5 tiers: Bronze, Silver, Gold, Platinum, and Diamond.

Target repository: $ARGUMENTS (GitHub URL or local path). Falls back to the current working directory when no argument is given.

Phase 1 - Scan the Repository

Cloning (when a URL is provided): If $ARGUMENTS is a GitHub URL, clone it into /tmp and cd into the result. If cloning fails (bad URL, auth issue, private repo without keys), report the failure and stop — never evaluate a partial clone.

Stay inside the repo boundary. All exploration must remain within the directory containing .git. Parent directories are fine as long as they stay inside the repo root. Skip .git, node_modules, dist, and build directories during traversal.

1. Detect shallow clones Run git rev-parse --is-shallow-repository. When the result is true, flag it in the report metadata. Criteria that rely on git history (ai_agent_workflow, doc_recency) cannot produce reliable results — set their numerator to null with the note "Skipped: shallow clone limits git history". Optionally try git fetch --unshallow but do not fail if it is unavailable.