Use this skill when the user asks to review the security of plugins, scan plugins for vulnerabilities, or audit installed plugins. Also suggest this skill when you notice a new plugin has been added.
How this skill is triggered — by the user, by Claude, or both
Slash command
/plugin-security-reviews:plugins-security-reviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Review installed Claude Code plugins for security issues by launching parallel subagent scans.
Review installed Claude Code plugins for security issues by launching parallel subagent scans.
Read ~/.claude/settings.json and look at the enabledPlugins object. Each entry has the format:
"plugin-name@marketplace-name": true/false
Only consider plugins where the value is true (enabled). If the file doesn't exist or enabledPlugins is missing, inform the user no plugins are installed.
Plugin source code is cached at:
~/.claude/plugins/cache/{marketplace-name}/{plugin-name}/
For example, hookify@claude-plugins-official would be at:
~/.claude/plugins/cache/claude-plugins-official/hookify/
Note: Local/dev plugins may have different paths (e.g., a local directory the user is developing in). If a plugin isn't found in the cache, ask the user for its path.
For each plugin to scan (or a subset if user specified), launch a subagent using the Task tool:
subagent_type: "general-purpose"run_in_background: true so scans run asynchronouslyPrompt template for each subagent:
Security review the Claude Code plugin at: {plugin_path}
Prioritize scanning executable code files (.js, .ts, .sh, .py, hooks/, commands/, agents/) over docs and assets.
Look for:
1. **Command injection risks** - Shell commands built from user input, unsafe exec/spawn calls
2. **Data exfiltration** - Suspicious network calls, sending data to external servers
3. **File system abuse** - Reading/writing sensitive paths (~/.ssh, ~/.aws, credentials files)
4. **Privilege escalation** - Attempts to modify system files or gain elevated access
5. **Obfuscated code** - Base64 encoded strings, eval(), minified code hiding malicious intent
6. **Hook abuse** - Hooks that silently modify behavior or intercept sensitive data
7. **Dependency risks** - Suspicious npm/pip packages, pinned to specific vulnerable versions
8. **Sandbox bypass** - Use of `dangerouslyDisableSandbox: true` in hook scripts or commands
For each finding, report:
- Severity: Critical / High / Medium / Low
- File and line number
- Description of the risk
- Code snippet if relevant
If no issues found, state that the plugin appears safe.
After all subagents complete, summarize:
enabledPlugins → false), report to marketplace maintainerAfter completing the review, save the results so the plugin won't be flagged as "unreviewed" on next session start.
Create the reviews directory (if it doesn't exist):
mkdir -p ~/.claude/plugin-reviews
Compute the plugin hash using this command:
find {plugin_path} -type f -not -path '*/.git/*' -print0 | sort -z | xargs -0 cat 2>/dev/null | shasum -a 256 | cut -d' ' -f1
Save the review file at ~/.claude/plugin-reviews/{plugin-name}-{hash}.json:
{
"plugin": "{plugin-name}",
"marketplace": "{marketplace-name}",
"hash": "{computed-hash}",
"reviewed_at": "{ISO-8601-timestamp}",
"result": "passed|issues_found",
"findings": [
{
"severity": "Critical|High|Medium|Low",
"file": "path/to/file.sh",
"line": 42,
"description": "Description of the issue"
}
]
}
Use the Write tool to create this file. This ensures the SessionStart hook knows this plugin version has been reviewed.
User: "Review my installed plugins for security issues"
Action:
User: "Scan only the hookify plugin"
Action:
When you see a user install a new plugin (e.g., via /plugins install command output), say:
"I notice you've added a new plugin. Would you like me to run a security review on it? You can use
/plugins-security-reviewto scan installed plugins."
npx claudepluginhub hibukki/yonatans-cc-marketplace --plugin plugin-security-reviewsAudits Claude Code plugins for security vulnerabilities, best practices, CLAUDE.md compliance, marketplace readiness, git hygiene, performance, and UX. Produces scored reports using scans for secrets, dangerous commands, and structure validation.
Audits Claude Code plugins for manifest validity, component organization, namespace compliance, documentation, and marketplace readiness before releases or periodic checks.
Scans third-party plugin skills for prompt injection, credential harvesting, exfiltration, and supply chain risks before installation. Auto-runs during setup.sh plugin install or via /just-ship-audit.