Read AGENTS.md: it is the source of truth for this project's conventions, shared across all AI tools (Claude Code, Copilot, Codex). The detailed, path-scoped rules it points to live in .github/instructions/.
Claude Code skills live in .claude/skills/:
/init- turn this cloned skeleton into a named project, and manage its identity, features, and example sets (drivesnpm run init)./scaffold- add one feature (CPT, block, REST controller, CLI command, ...) vianpx wp-tooling add, TDD-first.
A cross-repo knowledge graph of this plugin plus wp-tooling, wp-primitives, and wp-shared-workflows is committed at graphify-out/graph.json (human summary in graphify-out/GRAPH_REPORT.md), built with the official Graphify utility.
Before reading source files to answer questions about architecture, symbols, call paths, or how the four repos fit together, query the graph first to save tokens:
- Use the
/graphifyskill, or the CLI:graphify query "<question>",graphify path "A" "B",graphify explain "<symbol>",graphify affected "<symbol>". - The graph is structural (tree-sitter, local, no API key). Read files only when the graph does not answer.
After code changes, refresh only your local copy with graphify update . and never commit it. Rebuilding the committed cross-repo baseline is a maintainer task; recipe in docs/internal/knowledge-graph.md.
Claude-specific notes:
- (none currently; keep Claude overrides here if they ever diverge from AGENTS.md)