Audit a named ClawHub skill or skill URL before installation by combining OpenClaw verification with bounded static analysis. Use when the user explicitly asks whether a skill is safe or requests a pre-install review; report evidence and uncertainty instead of treating a score as proof.
Security
Skill Auditor
Try itDeep audit for installed ClawHub skills — usage analysis, permission review, conflict detection
What it does
Deep audit for installed ClawHub skills — usage analysis, permission review, conflict detection
The skill document
Skill Auditor (skill-auditor)
Audits all installed skills in the local OpenClaw environment. Scans metadata, files, permissions, dependencies, and usage history. Generates a health score, identifies zombie / duplicative / over-privileged skills, and produces cleanup scripts.
Workflow
- Scan installation — list all skills under
~/.openclaw/skills/and installed packages. Read eachSKILL.mdfor: name, description, allowed-tools, user-invocable, dependencies, license. - Usage analysis — parse session logs or call-frequency data over a configurable window (default 30 days). Compute: total invocations, trend (increasing / stable / declining / zero), last-used timestamp. Flag zombies (no usage in N days).
- Security audit — for each skill, check:
- Permission reasonability: does the skill's
allowed-toolsmatch actual tool usage? Flag over-privileged skills (e.g. a greeting skill withexecaccess). - Sensitive tool access: identify skills with
exec,web_fetch,scriptaccess. - Excessive cross-skill access.
- Permission reasonability: does the skill's
- Conflict detection — pairwise comparison:
- Overlapping
description/name(duplicate functionality). - Conflicting tool or file namespace (two skills defining the same helper).
- Circular dependencies.
- Overlapping
- Health score (0-100) — five dimensions weighted equally:
- Activity (usage frequency / recency, 0-20)
- Security (permission score, 0-20)
- Stability (error rate / completeness, 0-20)
- Maintainability (doc quality, file count, size, 0-20)
- Docs (description clarity, sample prompts, 0-20)
- Optimization suggestions — per skill:
- Uninstall: zombie skills with no value
- Suspend: rarely used but potentially useful
- Permission downgrade: over-permissioned
- Merge: overlapping skills into one
- Upgrade: missing metadata or outdated format
- Cleanup script — generate a shell script to: uninstall selected skills, disable (rename to
.disabled), suspend (archive). Provide dry-run mode. - Output report — structured Markdown report with: executive summary, per-skill audit card (health score + flags), conflict matrix, cleanup script path. Optionally export as HTML dashboard.
Sample prompts
skill-auditor auditskill-auditor audit --days 90 --min-activity 3skill-auditor audit --focus securityskill-auditor cleanup --dry-run
Related skills
Automated security audit for AI agent skills. Use BEFORE installing any skill from ClawHub, GitHub, or other sources. Scans SKILL.md + all files for 30+ red...
Audit a local OpenClaw or ClawHub skill directory before installation, classify it as PASS, REVIEW, or BLOCK, check active-project impact, and optionally run...
ClawHub reputation checker for clawsec-suite. Adds a standalone reputation gate before guarded skill installation.
Audit a ClawHub publisher portfolio using live registry evidence. Invoke when deciding which skills to improve, merge, stop, or build next based on downloads, installs, search competition, and version health.
Audit an upstream agent skill, SKILL.md, skill repository, or lifecycle workflow before adapting it for ClawHub, Codex, Claude Code, or a public Skool skill...