[hydraxman]

Fixes #106.

sloplint . dist --json can report a clean result without scanning dist, while sloplint dist . --json finds its violations. The same argument-order regression affects build and coverage.

The realpath guard added in #103 marks implicitly skipped build directories as visited. Move the existing directory-exclusion check before that guard, so only eligible paths enter seen. The guard still runs before recursion, preserving the cycle/alias fix by @xunguangzlj-cloud.

Add a ten-case regression to the existing refcheck.mjs runner: parent-only traversal, each explicit build target, and both argument orders. It checks file counts, findings and exit statuses. The existing symlink-cycle fixture remains unchanged.

Verification

• Real CLI probe on current main: three failures and seven passing controls; patched: all ten pass. The same probe also passes before fix(sloplint): deduplicate directory symlinks before recursion #103, confirming the regression boundary.
• Full node skills/hallmark/scripts/refcheck.mjs: clean across 101 reference files, including the new regression and existing symlink fixture.
• All 50 themes validate; regenerating the theme index leaves no diff.
• node --test skills/hallmark/scripts/theme.test.mjs: one test passes.
• git diff --check: passes.

Locally verified on macOS with Node 24.14.1. Browser/render integration was not run; no lint rules or dependencies change. Hosted CI is pending.

Prepared with AI assistance and independently checked by a second agent against the final diff and real CLI. This is not a claim of human review; maintainer review remains pending.

[vercel]

@bryan is attempting to deploy a commit to the Together AI Team on Vercel.

A member of the Team first needs to authorize it.