careful
Safety guardrails for destructive commands. Use for: be careful, warn before destructive, safety mode. (gstack)
How do I install this agent skill?
npx skills add https://github.com/garrytan/gstack --skill carefulIs this agent skill safe to install?
- Gen Agent Trust Hubpass
This skill provides a safety layer for terminal use by intercepting Bash and PowerShell commands to warn users before they execute potentially destructive operations like 'rm -rf', 'git reset --hard', or SQL 'DROP TABLE'.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
/careful — Destructive Command Guardrails
Safety mode is now active. Every Bash and PowerShell 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.
GSTACK_STATE_ROOT=$(~/.claude/skills/gstack/bin/gstack-paths --get GSTACK_STATE_ROOT); : "${GSTACK_STATE_ROOT:?gstack-paths failed; reinstall with ./setup or /gstack-upgrade}"
mkdir -p "$GSTACK_STATE_ROOT"/analytics
echo '{"skill":"careful","ts":"'$(date -u +%Y-%m-%dT%H:%M:%SZ)'","repo":"'$(basename "$(git rev-parse --show-toplevel 2>/dev/null)" 2>/dev/null || echo "unknown")'"}' >> "$GSTACK_STATE_ROOT"/analytics/skill-usage.jsonl 2>/dev/null || true
What's protected
Each command in a chain is checked on its own (after &&, sudo, xargs,
inside bash -c).
| Asks before | Example |
|---|---|
rm -r of anything but scratch (see below) | rm -rf /var/data |
Deleting outside the session folder: rm, find's delete actions, rsync --delete, shred, unlink | rm -f ~/notes.txt |
DROP TABLE/DATABASE through a SQL client; SQL TRUNCATE | psql -c "DROP TABLE users" |
Force-push (-f, +ref, --force-with-lease); remote branch delete | git push -f origin feature |
reset --hard, checkout -f/-- paths, restore, clean -f when there is work to lose; stash drop, worktree remove --force | git clean -fdx |
History rewrites (filter-branch, BFG, reflog expire --expire=now) | |
Database resets (prisma migrate reset, dropdb, FLUSHALL, DELETE FROM with no WHERE) | |
Cloud deletes (terraform destroy, aws/gcloud/az delete; not dry runs) | |
kubectl delete, docker rm -f, docker system prune |
PowerShell and cmd (Windows)
The hook also checks the PowerShell tool (Claude Code's main Windows shell)
and any pwsh/powershell/cmd launched from Bash. Matching ignores case,
covers aliases in command position, accepts parameter prefixes (-r, -fo)
and strips cmd ^ escapes. All rows above apply there too.
| Pattern | Example |
|---|---|
Remove-Item/rm/ri/del/erase/rd/rmdir + -Recurse or -Force | gci | ri -r -fo |
cmd rd /s, rmdir /s, del /s, erase /s | cmd /c rd /s /q C:\proj |
Format-Volume, Clear-Disk, Clear-Content, [IO.Directory]::Delete, [IO.File]::Delete | Clear-Disk -Number 1 |
-EncodedCommand, iex, Start-Process of a shell, & $cmd (can't be inspected) | irm $url | iex |
Best-effort on PowerShell. String matching can't see a command built at
runtime. For a hard stop, add Claude Code permission deny rules, which parse
PowerShell and its aliases: "deny": ["PowerShell(Remove-Item *)"] in
.claude/settings.json. The hook runs through bash, so Windows needs Git Bash.
Safe exceptions
A recursive delete never asks when every target is scratch: under /tmp/ or
/var/tmp/, a $(mktemp) variable, a relative path after cd into scratch, a
cache (~/.cache, ~/.npm and the like), or a build artifact
(node_modules, dist, build, .next, .venv and the like). A path containing .. is never scratch.
How it works
The hook checks each command against the patterns above and returns
permissionDecision: "ask" (nested under hookSpecificOutput) with a reason.
You can always approve an ask and proceed.
HIGH tier (hard deny)
Denied, not asked: a recursive delete of /, ~, a system folder or
~/.ssh-style personal folder; a force-push to main, master or the
default branch (--force-with-lease is never HIGH); a disk wipe
(dd of=/dev/sda, mkfs). Best-effort, not a policy boundary: you can still
run the command yourself.
Project patterns (additive only)
Add warn rules — one POSIX ERE per line, # comments OK — in
~/.gstack/careful-patterns.txt (global) or
~/.gstack/projects/<slug>/careful-patterns.txt (per-project). Consulted
after the built-in families, so config can only ADD rules, never suppress a
baseline warning. Invalid regex lines are skipped.
Always on
./setup registers this check for every session (destructive_guard), so it
runs even when /careful is never loaded; /guard still adds the edit boundary.
Turn it off with gstack-config set destructive_guard off; /careful's own
hooks then last until the conversation ends.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/garrytan/gstack/careful">View careful on skillZs</a>