AGENTS

Instructions for AI coding agents in this repository. For project overview, start with README.md.

Project Context

Do not edit the bundled plugin skill first. Update .agents/skills/material/ and then sync the bundle with python scripts/sync_plugin_bundle.py.
Preserve the release model and validation flow documented in docs/release-policy.md, docs/release-checklist.md, and docs/github-release-flow.md.
Keep changes small and reviewable.
Run the smallest relevant validation for the changed layer.

Graphify is optional. Use it for architecture, dependency, module-relationship, or large repository navigation tasks; do not run heavy graph generation automatically.
Keep Graphify output narrow: prefer graphify query "<specific question>" --budget 1200, graphify explain "<node>", or graphify path "<A>" "<B>" before reading full reports or doing broad source searches.
If the graph answer is too broad, refine by feature, domain, package, path, or named component and rerun with a smaller budget; do not paste large raw Graphify output into final answers.
For web/plugin code, scope Graphify to source-only paths such as .agents/skills/material, plugin, mcp, scripts, and relevant docs.
Exclude web/build noise from Graphify corpora: node_modules, .next, dist, build, out, coverage, playwright-report, test-results, storybook-static, static/public asset dumps, generated files, screenshots, graphify-out, and local caches.
Run Graphify on the smallest package/subtree that matches the task; widen only after the scoped graph and direct source reads are insufficient.