Activate safety guards for the session. Use /careful before critical operations — blocks rm -rf, DROP TABLE, git push --force, kubectl delete namespace, docker system prune via PreToolUse hook.
Save session state before context gets full or before ending. Writes current task state, branch, and loaded docs to claude/tasks/todo.md so the next session can resume cleanly.
Semantic vector search over the codebase — find code by meaning, not exact names. Use when you need to locate implementations without knowing exact function/class names. Triggers on: "find code", "search codebase", "locate implementation", "what does X do".
Change risk analysis — detect blast radius, risk score, and breaking changes from git diffs. Use before any PR or merge. Crown jewel: detectchangestool. Triggers on: "review PR", "check impact", "blast radius", "what breaks", "safe to merge".
Structural graph tools for live codebase navigation — trace call paths, detect change blast radius, find dead code, query architecture. Use before reading files. 120x fewer tokens than file exploration.
Load all relevant claude/ knowledge files for a domain area. Use at session start or when switching to a new topic. Accepts a domain keyword like api, database, auth, build, security.
Verify that a symbol (field, enum, status code) exists across all layers of a monorepo. Use after adding new fields, enums, or codes to catch cross-layer gaps.
Root cause analysis for bugs and errors. Use when something is broken and the cause is not obvious. Applies the 5-step investigation method — observe, hypothesize, trace backwards, binary search, prove.
Manage dependencies and fix CVEs — check outdated packages, update a specific package, run security audit, explain why a package is installed, or deduplicate. Read claude/cve-policy.md for the decision tree.
Verify Claude configuration health — CLAUDE.md, settings.json, hooks, agents, skills, secrets scan, and MCP reachability. Reports ✅/⚠️/❌ for each check.
Issue triage — audit open issues, categorize, detect duplicates, cross-ref PRs, risk assessment, post comments. Args: "all" for deep analysis, issue numbers to focus (e.g. "42 57"), "fr" for French, no arg = audit only in English.