Kjell Tore Guttormsen
03b8885b6e
~/.claude/CLAUDE.md specifies English for code and documentation, Norwegian for dialog only. Norwegian had crept into surface text across v7.5-v7.7. Translated to English in eight surfaces. No scanner, hook, or behavior changes — purely surface text. - 18 skill commands: the HTML Report-step now reads "HTML report: [Open in browser]" instead of "HTML-rapport: [Åpne i nettleser]" - scripts/lib/report-renderers.mjs: key-stat labels, lede defaults, table headers, maturity-ladder descriptions, action-tier labels, clean buckets, dry-run/apply copy, and JS comments. Regex alternations /^high|^høy/ and /resolution|løsning/i preserved. - playground/llm-security-playground.html: same renderer changes mirrored bit-identical, plus playground-only UI strings (catalog, breadcrumb aria-label, theme toggle, builder-modal hint, guide-panel "no projects yet", delete confirmation, alert/copy). Demo-state fixture content for dft-komplett-demo preserved (intentional Norwegian persona). - agents/skill-scanner-agent.md + agents/mcp-scanner-agent.md: Generaliseringsgrense + Parallell Read-strategi sections translated to Generalization boundary + Parallel Read strategy. - README.md: playground architecture prose + Recent versions table (v7.5.0 — v7.7.1). - CLAUDE.md: v7.7.1 highlights translated, new v7.7.2 highlights added. - ../../README.md: llm-security v7.5.0 — v7.7.1 bullets. - ../../CLAUDE.md: llm-security catalog entry. - docs/scanner-reference.md: six runnable-examples table cells. - docs/version-history.md: new v7.7.2 entry. v7.5-v7.7 narrative sections left in original language (deferred per operator). - Version bumped 7.7.1 → 7.7.2 in package.json, .claude-plugin/plugin.json, README badge + Recent versions, CLAUDE.md header + state, docs/version-history.md, playground renderHome hardcoded string, root README + CLAUDE.md llm-security entries. Tests: 1820/1820 green. CLI smoke-test: 18/18 commandIds produce >138 KB self-contained HTML. Browser-dogfood verified. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
67 lines
3 KiB
Markdown
67 lines
3 KiB
Markdown
---
|
|
name: security:mcp-audit
|
|
description: Audit all installed MCP server configurations for security risks, trust verification, and permission analysis
|
|
allowed-tools: Read, Glob, Grep, Agent, Bash
|
|
model: sonnet
|
|
---
|
|
|
|
# /security mcp-audit [--live]
|
|
|
|
Full MCP server security audit — project-level and global. Add `--live` to also connect to running servers and scan live tool descriptions.
|
|
|
|
## Step 0: Parse Flags
|
|
|
|
If `$ARGUMENTS` contains `--live`, strip it and set `run_live_inspection = true`.
|
|
|
|
## Step 1: Discovery
|
|
|
|
Read MCP configs from: `.mcp.json`, `.claude/settings.json`, `claude_desktop_config.json`, `~/.claude/settings.json`, `~/.claude/mcp.json`, `~/.config/claude/mcp.json`.
|
|
|
|
For each server extract: name, transport (stdio/sse), command+args or URL, env var names (redact values), source origin. Report total count.
|
|
|
|
## Step 2: Scan
|
|
|
|
Spawn `subagent_type: "llm-security:mcp-scanner-agent"`, `model: "sonnet"`:
|
|
|
|
> Full MCP audit. Read: \<plugin-root\>/knowledge/mcp-threat-patterns.md
|
|
> Execute all 5 phases per server (tool descriptions, source code, dependencies, config, rug pull detection).
|
|
> Servers: \<discovered server list with name, type, command, source, env vars\>
|
|
> Return per-server trust rating + findings.
|
|
|
|
## Step 3: Report
|
|
|
|
Output: MCP Landscape Summary table (server, source, transport, trust rating, finding counts).
|
|
Overall risk: Low (all trusted) / Medium (cautious+high) / High (untrusted) / Critical (dangerous).
|
|
Group servers: Keep / Review / Remove immediately.
|
|
|
|
## Step 4: Live Inspection (only if `--live`)
|
|
|
|
Run: `node <plugin-root>/scanners/mcp-live-inspect.mjs "<cwd>"`
|
|
|
|
Parse JSON output. Append a **Live Inspection Results** section:
|
|
|
|
- Contact status per server (contacted / timed-out / skipped / failed)
|
|
- Live injection findings (sorted critical → info)
|
|
- Tool shadowing across servers
|
|
|
|
**Cross-reference escalation:** If a server was rated "Untrusted" or "Dangerous" in Step 2 AND has live injection findings → escalate to CRITICAL priority in the final report and highlight as "Confirmed active threat (static + live)".
|
|
|
|
## Step 5: HTML Report
|
|
|
|
After producing the markdown MCP audit report (Step 3 + optional Step 4 Live Inspection):
|
|
|
|
1. Compute a temp markdown path:
|
|
```bash
|
|
node -p "require('path').join(require('os').tmpdir(), 'sec-mcp-audit-' + Date.now() + '.md')"
|
|
```
|
|
2. Use the Write tool to save the **entire markdown report you just produced** (MCP Landscape Summary + per-server analysis + Live Inspection Results section if `--live` + Keep/Review/Remove groupings + overall risk) to that temp path. Verbatim.
|
|
3. Run the renderer:
|
|
```bash
|
|
node <plugin-root>/scripts/render-report.mjs mcp-audit --in "<temp-md-path>"
|
|
```
|
|
The CLI writes `reports/mcp-audit-<YYYYMMDD-HHmmss>.html` relative to CWD and prints `file:///abs/path.html` on stdout.
|
|
4. Append to your response (markdown link, no bare URL):
|
|
|
|
> **HTML report:** [Open in browser](file:///abs/path.html)
|
|
|
|
If the CLI exits non-zero, mention the error but do not block — the markdown report above is the primary deliverable.
|