System package maintenance for Ubuntu (apt), npm, brew, and OpenClaw skills. Conservative workflow with 2-day quarantine for non-security updates, automatic...
Security Analysis
medium confidenceThe skill mostly matches its stated purpose (conservative updater for apt/npm/brew/OpenClaw), but there are multiple mismatches and risky/incoherent items (sudoers vs code, undeclared binaries/env, Telegram/sending expectations) that you should understand before installing.
The project is plausibly a system updater for apt/npm/brew and OpenClaw skills — that capability explains most included code (apt_maint.py, pkg_maint.py, auto_review.py) and the documentation. However registry metadata (no required binaries / no required env vars) does not match SKILL.md which declares required binaries (apt-get, npm, brew, clawhub) and environment variables (SYS_UPDATER_BASE_DIR, SYS_UPDATER_STATE_DIR, SYS_UPDATER_LOG_DIR). The claimed behavior (auto-review, skill updates via ClawHub, Telegram reports) is consistent with the code, but the metadata omission is an incoherence that makes automated permission checks unreliable.
Runtime instructions and code perform system-level actions: run sudo apt-get update, run unattended-upgrade (applies security updates), simulate upgrades, parse /var/log/apt/history.log, and call external services (npm registry, Homebrew API, GitHub API). The SKILL.md also instructs creating a sudoers file to allow passwordless sudo for apt/unattended-upgrade. The skill also documents auto-updating OpenClaw skills (clawhub) and generating Telegram reports; however, no Telegram token or sending code is declared (the repo expects the caller to forward/report). The scope (reading system logs, running sudoed commands, and making network requests) is consistent with an updater but is sensitive — especially because the instructions include granting passwordless sudo for system update commands.
There is no external install step (instruction-only + bundled Python scripts). The code uses only Python stdlib and subprocess calls; nothing is downloaded from arbitrary URLs. That lowers install risk compared to downloadable binaries. The repo includes code files (not pure prose), so a code review is appropriate before use.
The skill requires elevated capability (passwordless sudo entries) which is proportionate for a tool that applies security updates, but there are notable mismatches: 1) The documented sudoers entries intentionally exclude apt-get install (and docs repeatedly say "No package installation/removal through sudo"), yet the code's apply_planned_apt_upgrades calls 'sudo apt-get install -y <packages>' — this will fail unless sudoers is expanded (or the code is changed), creating a dangerous gap between intended permissions and actual behavior. 2) SKILL.md mentions Telegram reporting and OpenClaw/clawhub usage, but no credential/environment variables (e.g., Telegram bot token or clawhub config) are declared in registry metadata. 3) The registry metadata lists no required env vars while the docs and code expect SYS_UPDATER_* env vars — an inconsistency that affects deploy/automation and least-privilege questions.
The skill does not set always:true, but it asks the operator to create a sudoers file granting passwordless execution of system update commands for a user. Modifying /etc/sudoers.d is a permanent, high-impact change and should be done deliberately. Combined with the code calling 'sudo apt-get install' (which is not listed in the recommended sudoers), this elevates risk: either the installer will need to expand sudoers (increasing privilege), or planned installs will silently fail. The skill does not attempt to modify other skills' configurations directly in the code shown, but it does intend to auto-update OpenClaw skills (clawhub) which means it can change other skill state when run.
Guidance
What to check before installing or running this skill: 1) Read the code. The repo includes scripts that will run sudo commands, parse apt logs, and perform network requests to GitHub/npm/Homebrew. If you can't review them yourself, run them in an isolated test VM first. 2) Sudoers mismatch: the docs intentionally restrict passwordless sudo to apt-get update, apt-get -s upgrade (simulation) and unattended-upgrade. But the code calls 'sudo apt-get install' to apply planned non-security updates. Decide which behavior you want: either (A) keep sudoers restrictive and remove/disable planned installs in the code, or (B) explicitly add apt-get install to sudoers after understanding the consequences. Never add broad sudo rights — add only the exact commands (absolute paths) you audited. 3) Paths and binaries: verify correct absolute paths in sudoers. The docs/examples reference both /usr/bin/unattended-upgrade and /usr/sbin/unattended-upgrade in different places — confirm the actual path on your distribution. 4) Telegram/OpenClaw behavior: the skill renders reports but does not appear to ship or require a Telegram bot token; cron/OpenClaw is expected to forward the report. If you expect the skill to push messages itself, look for code that stores or uses bot tokens (none declared). Similarly, auto-updating OpenClaw skills implies using clawhub — ensure you understand what clawhub does and that the binary/credentials are available and limited. 5) Network calls and rate limits: auto_review.py will query public APIs (GitHub/npm/brew). It uses unauthenticated GitHub requests (60/hr). If you need higher rate or authenticated checks, consider adding a token and auditing how it is stored. 6) Permissions to read system files: parsing /var/log/apt/history.log may require root privileges on some systems; the code handles PermissionError but will produce incomplete data if it cannot read logs. 7) Test in dry-run: use the provided --dry-run and --verbose modes. Run on a non-production host and inspect generated state/logs before enabling cron/systemd or changing sudoers. 8) Reconciliation: if you accept planned installs, add tests and an audit path to ensure planned installs cannot be abused to install arbitrary packages. Prefer whitelisting package names or requiring manual confirmation for potentially dangerous packages. If you provide updated metadata (declare required binaries and env vars) and either remove/disable the sudo apt-get install usage or explicitly include it in the documented sudoers with audited absolute paths, my confidence that this skill is coherent and low-risk would increase.
Latest Release
v1.1.0
Improve skill description/docs; daily report now explicitly includes installed npm/pnpm/brew updates from pkg_maint summary.
More by @Spiceman161
Published by @Spiceman161 on ClawHub
