Operate a terminal for coding and system tasks with scoped working directories, command-level permissions, visible logs, timeouts, and no silent privilege escalation.
Install
npx skillscat add molotroka123-cell/aimaxbossman/safe-terminal Install via the SkillsCat registry.
SKILL.md
Safe Terminal Operator
Use when an agent needs shell access.
Default execution tiers
Tier 1 — Project Sandbox
Default for coder agents.
- workdir mounted only
- no host filesystem outside approved roots
- network off unless task explicitly needs it
- no sudo/admin
- command timeout
- stdout/stderr logged
AUTO examples:
- git status / diff / log
- tests
- linters
- local build
- read-only package/version checks
Tier 2 — Project Host Terminal
Use only when sandbox cannot perform the task.
ASK:
- package installation
- starting/stopping host services
- Docker operations affecting persistent services
- git push
- commands writing outside project
Tier 3 — System Admin
Explicitly privileged.
Always ASK:
- sudo / administrator
- firewall
- drivers
- system services
- mounting disks
- user/account management
DENY by default:
- destructive disk commands
- credential dumping
- disabling endpoint security
- arbitrary deletion outside approved workspace
- force-push or destructive git history rewrites
Whole-computer access
Never equate "terminal access" with unrestricted filesystem access.
A task must receive approved roots, for example:
D:/Projects/**read/writeC:/Users/<user>/Documents/**read- system directories deny
Secrets such as .env, SSH keys, browser password stores and crypto-wallet data require separate explicit policy and should not be model-visible.
Every command record
Log:
- agent
- task/run
- cwd
- exact command
- start/end
- exit code
- timeout/kill state
The dashboard must support Stop/Kill for active commands.