Installs the sem skill into ~/.claude/skills/sem/ so the agent knows when and how to reach for sem (impact, context, orient, diff, blame, log) instead of grep for structural code questions.
Registers the sem MCP server (sem mcp) at user scope, so the sem_impact / sem_context / sem_entities / ... tools are available in every session.

Optional: a live sem badge in your statusline

npx @ataraxy-labs/sem-skill --badge

Adds a live badge to your Claude Code statusline. The moment your agent triggers sem you see an animated spinner with the entity being analyzed (⊕ sem ⠹ impact validateToken…), flipping to the result and running savings when it completes. It shows in real time: how many structural queries this session, the last command and the entity it analyzed, its latency, a sparkline of recent latencies, and a rotating stat (distinct entities analyzed, top command) (⊕ sem ×12 impact validateToken 9ms ▁▂▃▅▂ · 7 entities analyzed). It picks up sem whether the agent uses the MCP tools or the sem CLI, and falls back to recent activity so it never gets stuck on "idle". It is opt-in and non-destructive: it backs up your settings, and if you already have a statusline it leaves it untouched and just tells you how to add the badge yourself. To remove it, delete the statusLine key and the mcp__sem__.* and Bash PostToolUse entries from ~/.claude/settings.json.

Team presence (opt-in): if you are logged in (sem login) and create ~/.sem/team.json with {"share": true}, your sessions heartbeat which entity your agent is working on, and the statusline shows teammates live (👥 maya → resolve_ref · 2m) so parallel agents stop colliding. Nothing is shared unless you opt in.

Everything renders inside the session: sem MCP calls show as proper tool widgets with compact entity trees, and the statusline badge tracks live time/token savings. No extra processes needed. Optionally, for a dedicated terminal pane, the install also drops ~/.claude/sem-live.py:

python3 ~/.claude/sem-live.py

It redraws an ASCII blast-radius graph every time sem runs (the analyzed entity, its direct callers, transitive count), plus a savings meter — a running, honestly-estimated tally of the grep+read round-trips, time, and tokens sem saved you this session, and a lifetime counter that grows across sessions. The estimates are anchored to a measured benchmark and labelled ≈.

It's idempotent, re-run it any time. It needs the sem CLI on PATH (npm i -g @ataraxy-labs/sem or see the install docs); if sem isn't installed yet, the skill and MCP registration still go in and work once it is.

Restart the agent session afterward to load the MCP tools.

Skill content originally contributed by @linhlban150612 (sem PR #376).