<!-- AUTO-GENERATED from SKILL.md.tmpl — do not edit directly. Run: node scripts/gen-skill-docs.mjs -->

---

name: careful description: > Safety guardrails for destructive commands. Warns before rm -rf, DROP TABLE, force-push, git reset --hard, kubectl delete, and similar destructive operations. Triggers on "careful", "조심", "안전 모드", "prod mode", "safety mode", "프로덕션 모드", "be careful", "careful mode". allowed-tools:

• Bash
• Read hooks: PreToolUse:
  • matcher: "Bash" hooks:
    • type: command command: "bash ${CLAUDE_SKILL_DIR}/bin/check-careful.sh" statusMessage: "Checking for destructive commands..."

---

/careful — Destructive Command Guardrails

Safety mode is now active. Every bash command will be checked for destructive patterns before running. If a destructive command is detected, you'll be warned and can choose to proceed or cancel.

Preamble

Before executing this skill:

1. Load context from memory:

   memory_search(query: "{skill-relevant-query}", associative: true, limit: 5)
   memory_search(tag: "{skill-name}", limit: 3)
   

   Review returned memories for relevant past context, decisions, and patterns.

2. Check OMC state for active work:

   state_get_status()
   

   If conflicting active tasks exist, warn the user before proceeding.

3. Detect current branch (for git-related skills):

   git rev-parse --abbrev-ref HEAD 2>/dev/null || echo "not-a-git-repo"
   

4. Check proactive mode:

   state_read("occ-proactive")
   

   If "false": do NOT proactively suggest other OpenClaw-CC skills during this session. Only run skills the user explicitly invokes.

5. Log skill activation:

   memory_daily_log(type: "note", entry: "Skill activated: /{skill-name}")
   

What's protected

Pattern	Example	Risk
rm -rf / rm -r / rm --recursive	rm -rf /var/data	Recursive delete
DROP TABLE / DROP DATABASE	DROP TABLE users;	Data loss
TRUNCATE	TRUNCATE orders;	Data loss
git push --force / -f	git push -f origin main	History rewrite
git reset --hard	git reset --hard HEAD~3	Uncommitted work loss
git checkout . / git restore .	git checkout .	Uncommitted work loss
kubectl delete	kubectl delete pod	Production impact
docker rm -f / docker system prune	docker system prune -a	Container/image loss
DELETE FROM (without WHERE)	DELETE FROM users	Mass data deletion

Safe exceptions

These patterns are allowed without warning:

• rm -rf node_modules / .next / dist / __pycache__ / .cache / build / .turbo / coverage / .omc

How it works

The hook reads the command from the tool input JSON, checks it against the patterns above, and returns permissionDecision: "ask" with a warning message if a match is found. You can always override the warning and proceed.

Memory Integration

memory_daily_log(type: "note", entry: "Safety mode activated: /careful")

To deactivate, end the conversation or start a new one. Hooks are session-scoped.

Completion Status Protocol

Every skill must end with one of these status codes:

Code	Meaning	When to Use
DONE	All steps completed, evidence provided	Root cause found + fix verified, PR created, review finished
DONE_WITH_CONCERNS	Completed with warnings or caveats	Tests pass but coverage dropped, fix applied but can't fully verify
BLOCKED	Cannot proceed, requires user intervention	3 failed attempts, missing permissions, external dependency down
NEEDS_CONTEXT	Missing information to continue	Unclear requirements, need user clarification

Escalation Rules

1. 3-strike rule: After 3 failed attempts at any step, STOP and escalate to user. Do not continue guessing. Present what was tried and ask for direction.

2. Scope escalation: If fix/change touches 5+ files unexpectedly, pause and confirm with the user before proceeding.

3. Security uncertainty: If you are unsure about a security implication, STOP and escalate. Never guess on security.

4. Verification requirement: Never claim DONE without evidence.

   • "Should work" → RUN IT. Confidence is not evidence.
   • "Already tested earlier" → Code changed since. Test again.
   • "Trivial change" → Trivial changes break production.

Output Format

═══════════════════════════════════════
Status: {DONE | DONE_WITH_CONCERNS | BLOCKED | NEEDS_CONTEXT}
Summary: {one-line description of outcome}
Evidence: {test output, verification results, or blocking reason}
═══════════════════════════════════════