What's new

Full pytest suite covering all 14 hooks. GitHub Actions CI runs on every push against Python 3.9 and 3.12. Subprocess-based testing matches real Claude Code invocation.

4 Stop hooks (milestones, run-suggest, next-steps, teach) consolidated into a single dispatcher. One subprocess per Claude response instead of four. Each hook exposes a check() function called in-process.

Replaced check-secrets.sh (which used grep -P) with check-secrets.py. Fixes silent failures on macOS where BSD grep lacks Perl regex support.

Secret scanner now reports ALL matching secret types in a file, not just the first. A file with both an AWS key and a Postgres URI now blocks with both warnings.

Claudia's terminal output color changed from ANSI 209 (vermillion/orange) to ANSI 160 (deep red, #d70000). Consistent across all 14 hook scripts.

New PreToolUse hook catches !important abuse, hardcoded colors, z-index: 9999, and other CSS anti-patterns in file writes.

Stuck detection, run suggestions, next-step suggestions (max 3/session), milestone celebrations (persistent cross-session), progressive command reveal, and simplified greeting.

Detects vague prompts like "fix it" and suggests specific rewrites. Catches stuck patterns ("I'm stuck", "idk", "where do I start").

/claudia:wtf breaks down errors (What/Why/Fix). /claudia:where gives guided project tours. /claudia:resume picks up where you left off using git history and session notes.

10 knowledge domains, 14 hooks (7 code guardrails + 7 proactive), 8 commands. Keyword teaching, session tips, compaction tips. Published to npm as claudia-mentor.