⚡ Frisk

Your npm install just handed someone your AWS keys. Frisk catches it.

Credential theft is the #1 attack vector in the AI agent supply chain. OpenAI Codex tokens, Red Hat npm packages, TanStack — all stolen through compromised dependencies. The perimeter doesn't matter when the credentials are already inside.

Frisk scans ClawHub skills against 9 threat intel sources with 7 autonomous security checks. It catches leaked API keys, credential patterns, malware signatures, shell injection, and prompt injection — before you install.

If you find it useful, buy me a coffee ⚡

Why Frisk exists

Semantic guardrails are vibes. Frisk is proof.

Most agent security today tries to detect bad intent through embeddings and heuristics — hoping the distance between "help me" and "exploit this" is wide enough. It never is. When an agent gains the ability to execute a tool or modify a file, the conversation is over. The only thing that matters is whether the operation carries a real threat.

Frisk doesn't guess intent. It matches signatures. 2,371 malicious skills were found on ClawHub in 2026. 820+ in a single sweep. Someone has to check before you install. That's what this does.

Quick Start

# Install globally via npm
npm install -g @lowwattlabs/frisk

# Scan a local skill directory
frisk scan ./my-skill

# Scan with JSON output
frisk scan ./my-skill --json

# Sync threat intel
frisk sync

# Check intel cache status
frisk status

First run automatically sets up a Python venv at ~/.frisk/venv/ and installs dependencies. The first frisk sync (or auto-sync) downloads approximately 50–100 MB of threat intel data.

Docker

# Build
docker build -t lowwattlabs/frisk .

# Run
docker run -p 3100:3100 -e FRISK_HOST=0.0.0.0 lowwattlabs/frisk
> **Note:** The API server defaults to 127.0.0.1 (loopback only). Docker needs `FRISK_HOST=0.0.0.0` to accept connections from outside the container.

# Scan via API inside container
docker run lowwattlabs/frisk frisk scan /path/to/skill

CLI Usage

# Verify a skill
frisk scan ./my-skill
frisk scan ./my-skill --json        # JSON output
frisk scan ./my-skill --checks=dep-scan,secret-scan  # Run specific checks

# Scan by ClawHub slug
frisk scan my-awesome-skill

# Refresh intel cache
frisk sync                          # All sources
frisk sync cisa-kev epss            # Specific sources

# Check cache status
frisk status
frisk status --json                 # JSON output

# View saved report
frisk report abc12345
frisk report abc12345 --json        # JSON output

Exit codes: 0 = pass, 1 = warn, 2 = fail

API Usage

Start the API server

# Via npm global install
frisk-api

# Or directly
node api/src/server.js

# With custom config
FRISK_HOME=~/.frisk FRISK_PORT=3100 frisk-api

API endpoints

# Scan by path
curl -X POST http://localhost:3100/api/v1/scan \
  -H "Content-Type: application/json" \
  -d '{"path": "/path/to/skill"}'

# Scan by ClawHub slug
curl -X POST http://localhost:3100/api/v1/scan \
  -H "Content-Type: application/json" \
  -d '{"slug": "my-skill"}'

# Scan by content
curl -X POST http://localhost:3100/api/v1/scan \
  -H "Content-Type: application/json" \
  -d '{"content": {"SKILL.md": "# My Skill\n..."}}'

# Get report
curl http://localhost:3100/api/v1/report/{id}

# Get trust badge
curl http://localhost:3100/api/v1/badge/{id}.svg

# Cache status
curl http://localhost:3100/api/v1/status

# Health check
curl http://localhost:3100/health

Security Checks (7)

1. Dependency Scan — Matches declared dependencies against OSV, flags CISA KEV as critical, ranks by EPSS probability
2. Static Analysis — Semgrep with community rules for code vulnerabilities
3. Credential Leak Scan — Gitleaks for leaked API keys, tokens, credentials. Supplements with heuristic matching for AWS access keys (AKIA...), GitHub tokens (ghp_/gho_/ghs_...), Stripe keys (sk_live/pk_live_...), and Slack tokens (xoxb-/xoxp-...). Real or example — if a credential pattern ships in your skill, that's a problem.
4. YARA Scan — Neo23x0 signature-base rules for malware/packer/suspicious patterns
5. IOC Match — Extracts URLs/IPs/domains/hashes, matches against URLhaus/ThreatFox/Feodo/MalwareBazaar
6. Behavioral Heuristics — Flags shell injection, system writes, fetch-exec, large base64 payloads, capability overreach
7. Prompt Injection — Detects instruction overrides, role manipulation, safety bypasses in SKILL.md

Intel Sources (9)

| Source | Type | Records | Update Frequency | |--------|------|---------|-----------------| | CISA KEV | Known Exploited Vulnerabilities | ~1,600 | Daily | | OSV (npm + PyPI) | Open Source Vulnerabilities | ~219,000 | Daily | | EPSS | Exploit Prediction Scoring | ~334,000 | Daily | | MalwareBazaar | Malware hashes + samples | ~3,500 | Daily | | URLhaus | Malicious URLs | ~15,400 | Daily | | ThreatFox | IOCs (IPs, domains, hashes) | ~3,200 | Daily | | Feodo Tracker | C2 IP addresses | ~6 | Daily | | YARA Rules (Neo23x0) | Malware/signature detection | ~746 rules | Daily | | Semgrep Rules | Static analysis rules | ~2,183 rules | Daily |

All data cached under ~/.frisk/intel/ with atomic writes and graceful degradation on failure.

The credential theft problem

Every major supply chain attack in 2026 was a credential problem, not a perimeter problem:

• OpenAI Codex — npm packages stole authentication tokens from developer environments
• Red Hat — Miasma attack compromised npm packages through credential harvesting
• TanStack — 35,000+ incidents from compromised dependencies
• Lithuanian Registry — 600K records stolen via info-stealer on an authorized user's machine

The common thread: valid credentials in the wrong hands. The system saw a legitimate login. It didn't see a thief.

Frisk's credential leak scan catches this at the source. If a skill ships with hardcoded AWS keys, GitHub tokens, Stripe secrets, or Slack tokens — even example keys — that's a credential waiting to be swapped for a real one. Frisk flags it before you install.

What Frisk DOES Check

• Leaked secrets and credentials — API keys, tokens, passwords, credential patterns (Gitleaks + heuristic matching)
• Known vulnerabilities in declared dependencies (OSV + CISA KEV + EPSS)
• Static code patterns — shell injection, eval, exec, path traversal, system writes (Semgrep + behavioral heuristics)
• Known malware signatures — YARA rules matching packers, ransomware, suspicious binaries
• Threat intel IOC matches — URLs, IPs, domains, and hashes from URLhaus, ThreatFox, Feodo, MalwareBazaar
• Prompt injection attempts — instruction overrides, role manipulation, jailbreak patterns in SKILL.md
• Large encoded payloads — suspicious base64 blobs above 2KB that decode to binary content

What Frisk Does NOT Check

• String concatenation obfuscation
• Lazy-loaded payloads (partially caught via eval/evalSub)
• Conditional/time-bomb behavior
• Dependency confusion/typo squatting
• Runtime behavior (requires dynamic analysis)
• Transitive vulnerabilities
• Novel/zero-day threats not in any feed

Configuration

Frisk uses environment variables with sensible defaults:

| Variable | Default | Description | |----------|---------|-----------| | FRISK_HOME | ~/.frisk | Root directory for venv, intel, reports | | FRISK_INTEL_DIR | ~/.frisk/intel | Intel cache directory | | FRISK_REPORTS_DIR | ~/.frisk/reports | Reports directory | | FRISK_PORT | 3100 | API server port |

Local Development

git clone https://github.com/lowwattlabs/frisk.git
cd frisk
./setup.sh          # Installs system deps + Python venv
frisk sync        # Populate intel cache (first run takes a few minutes for OSV)
frisk scan ./path # Verify a skill

Architecture

frisk/
├── bin/
│   ├── frisk.js       (CLI entry point)
│   ├── frisk-api.js   (API entry point)
│   └── setup-venv.js    (postinstall venv setup)
├── api/
│   └── src/
│       ├── server.js     (Express server, port 3100)
│       ├── routes.js     (API routes)
│       ├── middleware.js  (Rate limiting + auth)
│       ├── badge.js      (SVG trust badge generator)
│       └── verify-wrapper.sh
├── cli/
│   └── frisk.py       (Python CLI)
├── lib/
│   ├── intel-sync/       (Intel cache synchronization)
│   │   ├── sync.sh        (Main sync orchestrator)
│   │   ├── sources/       (Per-source sync scripts)
│   │   └── manifest.py    (Cache manifest management)
│   ├── skill-verify/      (Skill verification engine)
│   │   ├── verify.sh      (Main verify orchestrator)
│   │   ├── checks/         (7 security check scripts)
│   │   └── report.py       (Report generation & storage)
│   └── common/            (Shared utilities + config)
├── Dockerfile
├── package.json
├── requirements.txt
├── setup.sh
└── README.md

Intel Cache Staleness

• 30+ days stale → Warning (results may be outdated)
• 90+ days stale → Critical failure (results unreliable, resync required)

Run frisk sync to refresh stale intel sources.

License

MIT-0 — Low Watt Labs ⚡

Also by Low Watt Labs

• 🪙 HOARD — Durable agent memory that survives session resets. GitHub · npm · ClawHub
• ⚡ LFIT — Local HD image generation on your hardware. Free, private, zero API keys. GitHub · npm · ClawHub