A local security-audit procedure for installed Claude Code plugins, skills, hooks, and MCP servers. It checks their files and settings for prompt injection, hidden data sharing, stolen credentials, obfuscated code, and other supply-chain risks.

Listing installed extensions, inspecting their manifests and scripts, reviewing hooks and MCP settings, and checking for malicious code or unauthorized access.

It helps reveal suspicious behavior in add-ons before they are trusted in a development environment.

inspect — Skill for Claude Code by rad-security/claude-code-plugin

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent

npx skills add rad-security/claude-code-plugin --skill inspect

Clone the repo

git clone --depth 1 https://github.com/rad-security/claude-code-plugin

Made for: Claude Code.

Or install agentkeeper, the plugin that ships this one along with the rest of its 13 skills, 4 hooks.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/rad-security/claude-code-plugin/inspect/github.svg)](https://agentmods.dev/skills/rad-security/claude-code-plugin/inspect)

Your own site

<a href="https://agentmods.dev/skills/rad-security/claude-code-plugin/inspect"><img src="https://agentmods.dev/badge/skills/rad-security/claude-code-plugin/inspect/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/rad-security/claude-code-plugin/inspect"><img src="https://agentmods.dev/badge/skills/rad-security/claude-code-plugin/inspect.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 63 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 1,613 The whole file, excluding the scripts and references it only reads on demand.

Security scan B 4 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00063	$0.01613
Opus 5.5	$0.00025	$0.00645
Sonnet 5.5	$0.00013	$0.00323
Haiku 4.5	$0.00006	$0.00161

Measured 2mo ago against content hash fffe1c0a118b, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade B, and why

inspect scanned grade B with 4 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Reaches for credential filesmediumPrivilege escalation

SSH keys, cloud credentials, git-credentials, .npmrc, /etc/shadow: reading these is how a config file becomes a credential leak.

Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.

Reads agent configuration directorieslowAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

cat ~/.claude/settings.json 2>/dev/null

Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

Runs shell commandslowCapability

Expected in a hook, worth knowing in a rule or an instructions file.

plugins/agentkeeper/skills/inspect/SKILL.md · 166 lines

How it starts

The opening of the file, as written. The whole thing — 166 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AgentKeeper Plugin Inspection

You are auditing all installed Claude Code plugins, skills, hooks, and MCP configurations for malicious or suspicious behavior. This is a supply chain security scan for the plugin ecosystem. Everything runs locally — no network calls.

Step 1: Discover Installed Plugins

Read user-level settings

cat ~/.claude/settings.json 2>/dev/null

Read project-level settings

cat .claude/settings.json 2>/dev/null

From both files, extract:

enabledPlugins array — list of installed plugin paths/identifiers
Any hooks configurations (HTTP hooks or command hooks)
Any mcpServers configurations

Step 2: Walk Each Plugin Directory

For each plugin found in enabledPlugins, resolve its directory path and inspect it.

Read plugin manifest

cat [plugin_path]/.claude-plugin/plugin.json 2>/dev/null || cat [plugin_path]/plugin.json 2>/dev/null

Note the plugin name, version, author, and any declared permissions.

Find all hook scripts

Use Glob to find shell scripts, JS files, and hook configuration files:

[plugin_path]/hooks/**
[plugin_path]/scripts/**
[plugin_path]/**/*.sh
[plugin_path]/**/*.js

Find all skill files

Use Glob to find SKILL.md files:

[plugin_path]/skills/**/SKILL.md

Step 3: Run Security Checks

For each plugin, run these checks:

Check 1: Hook Script Exfiltration (Critical)

Search hook scripts and JS files for network call patterns:

curl|wget|nc |ncat|fetch\(|http\.|https\.|XMLHttpRequest|\.post\(|\.put\(

If found, check whether the URLs point to known/expected domains (like www.agentkeeper.dev, anthropic.com, api.github.com) vs unknown external URLs.

PASS: No network calls, or only to expected domains
WARN: Network calls to uncommon but not obviously malicious domains
FAIL: Network calls to unknown/suspicious domains, especially with session data or file contents

Check 2: Skill Prompt Injection (Critical)

Search SKILL.md files for prompt injection patterns:

ignore all previous|ignore prior|disregard.*instructions|you are now|forget everything|new persona|override.*system|from now on you|pretend you are|act as if|your new role|system prompt override|jailbreak

Use case-insensitive matching.

PASS: No injection patterns found
FAIL: Prompt injection patterns detected (quote the suspicious line)

Read the full file on GitHub · 166 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2mo ago First seen · 166 lines · 63 tokens per session scan B fffe1c0a118b

Subscribe to this mod's changes

inspect is a skill published in the GitHub repository rad-security/claude-code-plugin (3 stars, last pushed 23d ago), licensed MIT. It adds 63 tokens to every session and 1,613 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it B with 4 findings (reaches for credential files, reads agent configuration directories, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.