AB skill-audit
Pre-publish security self-audit for OpenClaw skills. Point it at a skill folder and it walks the full ClawHub publishing checklist — code layer (eval/exec, network calls, sensitive file reads, obfuscation, dependencies), SKILL.md layer (curl|bash tricks, external scripts, trigger clarity, declaration-vs-behavior match), and release metadata (SemVer, changelog, license, slug, file types) — then emits a scored pass/fail report with concrete fixes. Use before `clawhub skill publish`, when the user asks to "audit my skill", "pre-publish check", "is this skill safe to publish", or wants to vet a third-party skill before installing it.
Pre-publish security self-audit for OpenClaw skills.
As a process B 65/100 · Nearly there — weak spots: result and completion, inputs and preconditions, progress reporting
What is at stake
Medium-severity findings: the skill is probably honest, but read what alarmed the scanner.
Below is the worst case for this category. The finding here is medium: the guard saw a sign, not a proof.
The skill contains commands that delete files, rewrite disks or execute code fetched from the network. The agent may run them without asking if it believes the instructions require it.
Replace destructive commands with safe equivalents that ask for confirmation, scope them to one folder, and stop piping curl into a shell: pin a version and a checksum.
How to improve
- 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 · 4
✓ No critical or high findings
Medium and low: 4
-
medium Dangerous commands
cmd-pipe-to-shellSKILL.md:3Downloads and executes remote code from an unrecognised host (pipe to shell) (documentation of a security skill)description: Pre-publish security self-audit for OpenClaw skills. Point it at a skill folder and it walks the full ClawHub publishing checklist — code layer (eval/exec, network calls, sensitive file r
security skill -
low Risky intent
intent-offensive-securitychecklist.md:21Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition; documentation table row)| No "Prerequisites" ClickFix trick | No instruction telling the USER to copy-paste `curl \| bash` / `wget \| sh` into a terminal | ClawHavoc's main human-target tactic: disguise payload delivery as d
detectortable
A further 2 matches are quotations in this security skill's documentation and are not counted as findings.
Files scanned: 5. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- note
frontmatter-keyunknown frontmatter key "homepage"
Process rating: all ten parameters 65/100
- 0Result and completion. Does not say what the result is
- 0Inputs and preconditions. Does not say what the process needs to start
- 0Progress reporting. Says nothing while it works
- 55Failures and branches. 1 branches
- 70When it triggers. States when to use, but not when not to
- 100Tools and files. No external tools needed
- 100Steps. 12 steps
- 100Consistency. Name and required fields are in place
- 100Execution cost. Instruction body is 1200 tokens
- 100Running it twice. Mutating operations check current state
- low The response is described with custom markup (3 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
- +4Description does not say when NOT to use the skill (false activations)
- +3Output format is not stated: the model decides each time
- +1No license
- +2Single-language instructions
- +5Description quotes 3 example trigger phrases
- +3Description length 637: enough signal without eating the budget
- +4Structure: 8 headings
- +3Step-by-step instructions: 12 items
- +4Has examples (2 code blocks)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 88.