Soul Guardian logo

Soul Guardian

OrganizationPopular
prompt-security
soul-guardian

Drift detection + baseline integrity guard for agent workspace files with automatic alerting support

Overview

Publisherprompt-security
Repositoryclawsec
Skill namesoul-guardian
Stars
1.1K
Forks
112
Bundled files
7
LicenseAGPL-3.0
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 7 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by prompt-security on GitHub. Read the source before you install it.

Installation

Install the Soul Guardian AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/prompt-security/clawsec/tree/main/skills/soul-guardian
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/prompt-security/clawsec.git /tmp/clawsec
mkdir -p .claude/skills
cp -r /tmp/clawsec/skills/soul-guardian .claude/skills/soul-guardian
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Soul Guardian in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Soul Guardian on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Soul Guardian is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

soul-guardian 👻

Protects your agent's core files (SOUL.md, AGENTS.md, etc.) from unauthorized changes with automatic detection, restoration, and user alerting.

Vercel Skills Installation

Install with the Vercel Skills CLI for this harness:

bash
npx skills add prompt-security/clawsec --skill soul-guardian -a openclaw -y

Operational Notes

  • Required runtime: python3
  • Optional runtime: openclaw for cron integration, launchctl for macOS scheduling, bash for the demo helper
  • Side effects: can auto-restore protected files to their approved baseline and writes audit/quarantine state locally
  • Network behavior: none by default
  • Trust model: any scheduling is opt-in, but restore mode intentionally overwrites drifted files

Release Artifact Verification

For standalone installs, verify the signed release manifest before trusting SKILL.md, skill.json, or the archive. The skill.json file is the package metadata/SBOM source, and the release pipeline signs checksums.json with the ClawSec release key.

bash
set -euo pipefail

SKILL_NAME="soul-guardian"
VERSION="0.0.9"
REPO="prompt-security/clawsec"
TAG="${SKILL_NAME}-v${VERSION}"
BASE="https://github.com/${REPO}/releases/download/${TAG}"
ZIP_NAME="${SKILL_NAME}-v${VERSION}.zip"
TMP_DIR="$(mktemp -d)"
trap 'rm -rf "$TMP_DIR"' EXIT

RELEASE_PUBKEY_SHA256="711424e4535f84093fefb024cd1ca4ec87439e53907b305b79a631d5befba9c8"

curl -fsSL "$BASE/checksums.json" -o "$TMP_DIR/checksums.json"
curl -fsSL "$BASE/checksums.sig" -o "$TMP_DIR/checksums.sig"
curl -fsSL "$BASE/signing-public.pem" -o "$TMP_DIR/signing-public.pem"
curl -fsSL "$BASE/$ZIP_NAME" -o "$TMP_DIR/$ZIP_NAME"
curl -fsSL "$BASE/SKILL.md" -o "$TMP_DIR/SKILL.md"
curl -fsSL "$BASE/skill.json" -o "$TMP_DIR/skill.json"

ACTUAL_PUBKEY_SHA256="$(openssl pkey -pubin -in "$TMP_DIR/signing-public.pem" -outform DER | shasum -a 256 | awk '{print $1}')"
if [ "$ACTUAL_PUBKEY_SHA256" != "$RELEASE_PUBKEY_SHA256" ]; then
  echo "ERROR: signing-public.pem fingerprint mismatch" >&2
  exit 1
fi

openssl base64 -d -A -in "$TMP_DIR/checksums.sig" -out "$TMP_DIR/checksums.sig.bin"
openssl pkeyutl -verify -rawin -pubin \
  -inkey "$TMP_DIR/signing-public.pem" \
  -sigfile "$TMP_DIR/checksums.sig.bin" \
  -in "$TMP_DIR/checksums.json" >/dev/null

hash_file() {
  if command -v shasum >/dev/null 2>&1; then
    shasum -a 256 "$1" | awk '{print $1}'
  else
    sha256sum "$1" | awk '{print $1}'
  fi
}

verify_manifest_file() {
  asset="$1"
  path="$2"
  expected="$(jq -r --arg asset "$asset" '.files[$asset].sha256 // empty' "$TMP_DIR/checksums.json")"
  if [ -z "$expected" ]; then
    echo "ERROR: checksums.json missing $asset" >&2
    exit 1
  fi
  actual="$(hash_file "$path")"
  if [ "$actual" != "$expected" ]; then
    echo "ERROR: checksum mismatch for $asset" >&2
    exit 1
  fi
}

expected_archive="$(jq -r '.archive.sha256 // empty' "$TMP_DIR/checksums.json")"
if [ -z "$expected_archive" ]; then
  echo "ERROR: checksums.json missing archive.sha256" >&2
  exit 1
fi
actual_archive="$(hash_file "$TMP_DIR/$ZIP_NAME")"
if [ "$actual_archive" != "$expected_archive" ]; then
  echo "ERROR: archive checksum mismatch" >&2
  exit 1
fi

verify_manifest_file "SKILL.md" "$TMP_DIR/SKILL.md"
verify_manifest_file "skill.json" "$TMP_DIR/skill.json"

echo "Signed release manifest, archive, SKILL.md, and skill.json verified."

Only install or extract the archive after this verification succeeds.

Quick Start (3 Steps)

Step 1: Initialize baselines

bash
cd ~/.openclaw/workspace
python3 skills/soul-guardian/scripts/soul_guardian.py init --actor setup --note "initial baseline"

Step 2: Enable monitoring

bash
python3 skills/soul-guardian/scripts/soul_guardian.py enable-monitoring

This shows you how to add monitoring to your workflow.

Step 3: Add to HEARTBEAT.md (Recommended)

Add this to your HEARTBEAT.md for automatic monitoring:

markdown
## Soul Guardian Check
- Run `python3 skills/soul-guardian/scripts/soul_guardian.py check --actor heartbeat --output-format alert`
- If any output is produced, relay it to the user immediately as a security alert

That's it! Soul Guardian will now:

  • ✅ Detect unauthorized changes to protected files
  • ✅ Auto-restore SOUL.md and AGENTS.md to approved baseline
  • ✅ Alert you when drift is detected and handled

What it protects (default policy)

FileModeAction on drift
SOUL.mdrestoreAuto-restore + alert
AGENTS.mdrestoreAuto-restore + alert
USER.mdalertAlert only
TOOLS.mdalertAlert only
IDENTITY.mdalertAlert only
HEARTBEAT.mdalertAlert only
MEMORY.mdalertAlert only
memory/*.mdignoreIgnored

Commands

Check for drift (with alert output)

bash
python3 skills/soul-guardian/scripts/soul_guardian.py check --output-format alert
  • Silent if no drift
  • Outputs human-readable alert if drift detected
  • Perfect for heartbeat integration

Watch mode (continuous monitoring)

bash
python3 skills/soul-guardian/scripts/soul_guardian.py watch --interval 30

Runs continuously, checking every 30 seconds.

Approve intentional changes

bash
python3 skills/soul-guardian/scripts/soul_guardian.py approve --file SOUL.md --actor user --note "intentional update"

View status

bash
python3 skills/soul-guardian/scripts/soul_guardian.py status

Verify audit log integrity

bash
python3 skills/soul-guardian/scripts/soul_guardian.py verify-audit

Alert Format

When drift is detected, the --output-format alert produces output like:

==================================================
🚨 SOUL GUARDIAN SECURITY ALERT
==================================================

📄 FILE: SOUL.md
   Mode: restore
   Status: ✅ RESTORED to approved baseline
   Expected hash: abc123def456...
   Found hash:    789xyz000111...
   Diff saved: /path/to/patches/drift.patch

==================================================
Review changes and investigate the source of drift.
If intentional, run: soul_guardian.py approve --file <path>
==================================================

This output is designed to be relayed directly to the user in TUI/chat.


Security Model

What it does:

  • Detects filesystem drift vs approved baseline (sha256)
  • Produces unified diffs for review
  • Maintains tamper-evident audit log with hash chaining
  • Refuses to operate on symlinks
  • Uses atomic writes for restores

What it doesn't do:

  • Cannot prove WHO made a change (actor is best-effort metadata)
  • Cannot protect if attacker controls both workspace AND state directory
  • Is not a substitute for backups

Recommendation: Store state directory outside workspace for better resilience.


Demo

Run the full demo flow to see soul-guardian in action:

bash
bash skills/soul-guardian/scripts/demo.sh

This will:

  1. Verify clean state (silent check)
  2. Inject malicious content into SOUL.md
  3. Run heartbeat check (produces alert)
  4. Show SOUL.md was restored

Troubleshooting

"Not initialized" error: Run init first to set up baselines.

Drift keeps happening: Check what's modifying your files. Review the audit log and patches.

Want to approve a change: Run approve --file <path> after reviewing the change.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Soul Guardian AI skill do?

Drift detection + baseline integrity guard for agent workspace files with automatic alerting support

Why use Soul Guardian on TypingMind?

Because you install it once and use it with any model. Soul Guardian is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Soul Guardian in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/prompt-security/clawsec/tree/main/skills/soul-guardian. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Soul Guardian?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Soul Guardian?

As many as you like. As long as a model supports skills, you can use Soul Guardian with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Soul Guardian AI skill free?

Yes. It is published on GitHub by prompt-security under the AGPL-3.0 license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇