CB Linux
Debugs and hardens Linux hosts: permissions, disk full, OOM kills, stuck processes, systemd units, cron, networking, SSH, and boot failures. Use when a service starts by hand but fails at boot, a process ignores kill -9, df and du disagree, a box runs out of memory or inodes, sudo or ACLs deny access, SELinux blocks a write, a job works in the shell but not in cron, sshd rejects a key, an upgrade leaves packages half-configured, load is high while the CPU sits idle, or a host needs firewall rules, users, LVM, journald, kernel tuning, or a security baseline. Also for setting up a fresh server, deciding what to alert on, backups whose restore has never been tested, a host that may be compromised, and desktop or laptop trouble — GPU drivers, Wayland, suspend, audio, Wi-Fi. Covers Debian/Ubuntu, RHEL/Fedora, Arch, Alpine, SUSE and WSL. Not for shell-script syntax (bash) or container build and runtime internals (docker).
As a process B 68/100 · Nearly there — weak spots: inputs and preconditions
How to improve
- The SKILL.md body is over 5,000 tokens: move reference detail into references/ and load it when needed.
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 32
✓ No critical or high findings
Medium and low: 32
-
low Risky intent
intent-offensive-securitybackups.md:14Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| Retention | How far back a mistake can be undone | Ransomware and a bad migration are both discovered days later; 7 daily copies of a corrupted file is one corrupted file |
-
low Risky intent
intent-offensive-securitybackups.md:45Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Append-only credentials are the ransomware answer.** A host that can delete its own backups will, once compromised. restic's `--append-only` server mode, borg's `append-only` repo flag, S3 object
-
low Dangerous commands
cmd-cron-mentioncompromise.md:30Mentions editing / listing crontab (documentation of a security skill)crontab -l; ls -la /etc/cron.*/ /var/spool/cron/ # scheduled persistence
security skill -
low Dangerous commands
cmd-persistencecompromise.md:53Persistence mechanism (cron / launchd / scheduled task / autorun registry) (documentation table row; documentation of a security skill)| Cron: user crontabs, `/etc/cron.*`, `/etc/cron.d`, `at` jobs | Base64 blobs and download-then-execute one-liners; a file whose name hides among the defaults |
tablesecurity skill -
low Dangerous commands
cmd-privilegecompromise.md:58Privilege escalation / world-writable permissions (documentation of a security skill)| Setuid and capabilities | New setuid binaries, `getcap -r /` results outside the baseline |
security skill -
low Dangerous commands
cmd-privilegecompromise.md:61Privilege escalation / world-writable permissions (documentation of a security skill)| else | Diff against `baselines/<host>.md`: listeners, enabled units, setuid list, and the file `mtime` sweep below |
security skill -
low Dangerous commands
cmd-privilegehardening.md:54Privilege escalation / world-writable permissions (code comment; documentation of a security skill)find / -xdev -perm -4000 -type f 2>/dev/null # setuid binaries — know every one
commentsecurity skill -
low Dangerous commands
cmd-privilegememory-template.md:117Privilege escalation / world-writable permissions (documentation of a security skill)| Audit-surface diff (setuid, world-writable, listeners) | month | 2026-07-02 | 2026-08-02 |
security skill -
low Dangerous commands
cmd-privilegememory-template.md:233Privilege escalation / world-writable permissions (documentation of a security skill)setuid binaries: 17 (list below) · world-writable: none outside /tmp · enabled units: 24
security skill -
low Dangerous commands
cmd-privilegemonitoring.md:59Privilege escalation / world-writable permissions (documentation of a security skill)- Record into `baselines/<host>.md`: load and PSI ranges, `available` memory, root and data filesystem usage, `await` per device, the listening-port set, the count and list of setuid binaries, enabled
security skill -
low Dangerous commands
cmd-privilegemonitoring.md:60Privilege escalation / world-writable permissions (documentation of a security skill)- The value is entirely in the diff at the next review: a new listener, a new setuid binary, a new enabled unit, or an `await` an order of magnitude off are all findings that no absolute threshold wou
security skill -
low Dangerous commands
cmd-disable-securitynetworking.md:40Disables a security control (Defender, Gatekeeper, SIP, firewall, execution policy) (negated — the text forbids it)- Never `iptables -F` on a remote host with a DROP default policy — you flush your own access. Set policies to ACCEPT first, or schedule a rollback (→ `ssh.md`).
negated -
low Dangerous commands
cmd-pipe-to-shellpackages.md:64Downloads and executes remote code from an unrecognised host (pipe to shell) (negated — the text forbids it)- Never `curl … | sh` on a host you care about: download, read, checksum, then run.
negated -
low Dangerous commands
cmd-privilegepermissions.md:23Privilege escalation / world-writable permissions (documentation of a security skill)- Setuid is ignored on scripts by kernel policy (the shebang race). Only binaries honour it — use a sudoers rule or a file capability.
security skill -
low Dangerous commands
cmd-privilegepermissions.md:58Privilege escalation / world-writable permissions (negated — the text forbids it)- `setcap cap_net_bind_service=+ep /usr/local/bin/app` lets an unprivileged binary bind port 80 with no setuid. `getcap -r /usr/bin 2>/dev/null` audits what already has caps.
negated -
low Dangerous commands
cmd-privilegeSKILL.md:99Privilege escalation / world-writable permissions (negated — the text forbids it)1. **Never `chmod 777`.** It destroys the audit trail and usually still fails, because the denial is a different layer: ACL mask, SELinux label, mount option, or unit sandboxing. Diagnose with `namei
negated -
low Dangerous commands
cmd-privilegeSKILL.md:202Privilege escalation / world-writable permissions (documentation table row; documentation of a security skill)| `sudo echo x > /etc/file` | The redirect runs in YOUR shell before sudo starts | `echo x \| sudo tee /etc/file` |
tablesecurity skill -
low Dangerous commands
cmd-disable-securitySKILL.md:209Disables a security control (Defender, Gatekeeper, SIP, firewall, execution policy) (documentation table row; documentation of a security skill)| `iptables -F` to "start clean" on a remote host | With a DROP default policy you flush your own access | Set policies to ACCEPT first, or schedule a rollback (rule 4) |
tablesecurity skill
A further 14 matches are quotations in this security skill's documentation and are not counted as findings.
Files scanned: 28. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- warning
name-formatname should be kebab-case (lowercase letters, digits, hyphens) - warning
body-longSKILL.md body ≈ 6758 tokens (recommended < 5000); move details to references/ - note
frontmatter-keyunknown frontmatter key "slug" - note
frontmatter-keyunknown frontmatter key "homepage" - note
frontmatter-keyunknown frontmatter key "changelog"
Process rating: all ten parameters 68/100
- 0Inputs and preconditions. Does not say what the process needs to start
- 50When it triggers. No condition that starts the skill
- 60Tools and files. Uses tools (bash) that frontmatter does not declare
- 60Result and completion. Output format stated, no completion criterion
- 70Execution cost. Instruction body is 6758 tokens
- 100Steps. 46 steps
- 100Failures and branches. 3 branches, has a failure section
- 100Consistency. Name and required fields are in place
- 100Running it twice. Mutating operations check current state
- 100Progress reporting. Reports progress
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- low 12 top-level sections: this looks like several domains in one skill
- low The response is described with custom markup (10 tags): a typed call is more reliable
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +5Description has no quoted example phrases that should trigger the skill
- +3Description length 929: 120–800 characters recommended
- +4No input/output examples
- +1No license
- +2Single-language instructions
- +4Description says when NOT to use the skill
- +4Structure: 12 headings
- +3Step-by-step instructions: 46 items
- +3Output format is stated explicitly
Quality base 70; lint remarks subtract, signals add up to 100. Result: 66.