claudit-sec CLAUDE.md

Project instructions for a macOS security-audit tool that checks Claude Desktop and Claude Code settings, extensions, permissions, scheduled tasks, and runtime state.

In plain words
What is it for?
Maintaining the audit script, its data collectors, finding summaries, and ASCII, HTML, or JSON reports.
Why use it?
They give a coding agent the project structure and data-collection rules needed to work on the audit tool without guessing.

Instructions file

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.

agentmods
npx agentmods add instructions/harmonicsecurity/claudit-sec/claude-md
Clone the repo
git clone --depth 1 https://github.com/HarmonicSecurity/claudit-sec
Per session 2,074 This file is loaded in full into every session.
When invoked 2,074 The same file — it is already loaded in full.
Security scan B 3 findings. Scan, not verified.
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.

ModelPer sessionOnce invoked
Fable 5 $0.02074 $0.02074
Opus 5 $0.01037 $0.01037
Sonnet 5 $0.00415 $0.00415
Haiku 4.5 $0.00207 $0.00207

Measured 3d ago against content hash 94d8cde7e650, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

claudit-sec CLAUDE.md scanned grade B with 3 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 3d 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.

Asks for rootlowPrivilege escalation

A mod that escalates privileges can change anything on the machine, not only the project.

When run as root (uid 0) — which is how MDMs like FleetDM, Jamf, Mosyle, and CrowdStrike RTR execute scripts — the script automatically scans all users with Claude data. No flags needed.

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.

| Claude Code Settings | `~/.claude/settings.json` | Permission grants |

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

Enumerates other installed skillslowAgent snooping

Other skills' SKILL.md files reveal prompts, capabilities and secrets that should be invisible to peers.

~/.skills/skills/*/SKILL.md # Alt skills dir ~/.claude/skills/*/SKILL.md # Claude Code user skills

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

CLAUDE.md · 162 lines

How it starts

The opening of the file, as written. The whole thing — 162 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDIT-SEC — Claude Security Audit Tool

Project Overview

CLAUDIT-SEC is a read-only, single-file security audit tool for macOS that inspects Claude Desktop and Claude Code configuration, scheduled tasks, extensions, plugins, skills, permissions, and runtime state.

  • claude_audit.sh — Zsh (~1900 lines, requires jq) — designed for MDM/CrowdStrike RTR style deployment

This project is built and maintained using Claude Code. This file provides architecture context and development guidelines that Claude Code uses when working on the codebase.

Architecture

The script follows this structure:

  1. Preflight checks — OS validation, dependency checks
  2. User context detection — single user, explicit --user, or all-users scan
  3. Session directory discovery — finds local-agent-mode-sessions/org/user/ paths
  4. 14 data collectors — each reads specific config files and populates findings
  5. Finding aggregation — counts by severity (WARN, INFO, REVIEW)
  6. 3 output renderers — ASCII (ANSI color + Unicode tables), HTML (dark theme), JSON (SIEM-ready)

Data Collectors

Collector Source Files Key Findings
Desktop Settings claude_desktop_config.json → preferences keepAwakeEnabled, allowAllBrowserActions, menuBar, sidebar
Cowork Settings cowork_settings.json, config.json scheduledTasks, webSearch, networkMode
MCP Servers claude_desktop_config.json → mcpServers Server names, commands, env var keys
Plugins installed_plugins.json, remote manifest, marketplace cache installed/remote/cached plugins
Plugin Hooks hooks/hooks.json in plugin directories Shell commands on lifecycle events
Connectors local_*.json → remoteMcpServersConfig, .mcp.json web/desktop/not_connected, egressAllowedDomains, disabledMcpTools
Skills User skills (6 paths), installed plugin skills (3 paths) — see Key Paths SKILL.md frontmatter parsing
Scheduled Tasks scheduled-tasks.json Cron expressions with english translation
Extensions (DXT) extensions-installations.json Signature status, dangerous tools
Extension Settings Claude Extensions Settings/*.json Allowed directories
Blocklist extensions-blocklist.json Governance control presence
Claude Code Settings ~/.claude/settings.json Permission grants
Runtime pgrep, pmset, crontab, LaunchAgents Running processes, sleep assertions
Cookies Cookies, Cookies-journal Presence and permissions

Read the full file on GitHub · 162 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.

  1. 3d ago First seen · 162 lines · 2,074 tokens per session scan B 94d8cde7e650

Subscribe to this mod's changes

claudit-sec CLAUDE.md is an instructions file published in the GitHub repository HarmonicSecurity/claudit-sec (294 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 2,074 tokens to every session, about $0.0104 per session on Opus 5. A static security scan graded it B with 3 findings (asks for root, reads agent configuration directories, enumerates other installed skills). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.