@yourclaw/clawguard-cli

Command-line interface for the ClawGuard security scanner. Scan AI agent skills for prompt injection, secrets, malware, and permission abuse from your terminal.

Installation

# Global install
npm install -g @yourclaw/clawguard-cli

# Or run directly with npx
npx @yourclaw/clawguard-cli scan /path/to/skill

For local development as part of the ClawGuard project:

cd ../clawguard && make setup

Commands

clawguard scan <path>

Scan a skill directory for security issues.

# Basic scan (builtin patterns only, no external tools)
clawguard scan ./my-skill --builtin-only --skip-ai

# Full scan with all available tools
clawguard scan ./my-skill

# Output as SARIF for GitHub Code Scanning
clawguard scan ./my-skill --format sarif > results.sarif

# CI-friendly output with exit codes
clawguard scan ./my-skill --format ci

Options:

| Flag | Description | | ---- | ----------- | | --builtin-only | Skip external tools (gitleaks, semgrep, etc.) | | --skip-ai | Skip AI review for ambiguous findings | | --format <fmt> | Output format: terminal (default), json, sarif, markdown, ci |

Exit codes (CI format):

| Code | Meaning | | ---- | ------- | | 0 | Passed — no issues found | | 1 | Warning — medium-severity findings | | 2 | Blocked — critical/high-severity findings |

clawguard check <name>

Look up a skill in the ClawGuard trust registry.

clawguard check memory-manager
clawguard check code-formatter --registry https://clawguard.sh

Options:

| Flag | Description | | ---- | ----------- | | --registry <url> | Registry URL (default: https://clawguard.sh) |

clawguard doctor

Check which external scanning tools are available on your system.

clawguard doctor

Example output:

ClawGuard Doctor
════════════════

  gitleaks    ✓ installed
  semgrep     ✓ installed
  mcp-scan    ✗ not found
  npm         ✓ installed

Examples

Scan a local skill

$ clawguard scan ./my-memory-skill --builtin-only --skip-ai

  ClawGuard Scan Report
  ═════════════════════

  Skill:     my-memory-skill
  Status:    ✓ PASSED
  Score:     0 / 100
  Findings:  0

  Recommendation: Safe to install

Scan a skill with findings

$ clawguard scan ./suspicious-skill --builtin-only --skip-ai

  ClawGuard Scan Report
  ═════════════════════

  Skill:     suspicious-skill
  Status:    ✗ BLOCKED
  Score:     85 / 100
  Findings:  7

  CRITICAL  PI-001  Prompt injection: instruction override attempt
  CRITICAL  MAL-002 Reverse shell detected
  HIGH      MAL-003 Data exfiltration pattern
  HIGH      SEC-001 AWS access key detected
  ...

  Recommendation: Do not install

Use in CI/CD

# .github/workflows/scan.yml
- name: Scan skill
  run: npx @yourclaw/clawguard-cli scan ./skill --builtin-only --skip-ai --format ci

Development

# Install dependencies
npm install

# Build
npm run build

# Run locally (without building)
node bin/clawguard.js scan ../clawguard-rules/test-fixtures/malicious/data-exfiltration-skill --builtin-only --skip-ai

# Lint
npm run lint

Makefile

make install    # npm install
make build      # tsup build
make test       # run the demo scan
make lint       # biome check
make clean      # remove dist/ and node_modules/

Contributing

See the main CONTRIBUTING.md for guidelines. To work on the CLI specifically:

git clone https://github.com/yourclaw/clawguard.git && cd clawguard
make bootstrap    # sets up all repos
cd ../clawguard-cli
# make changes...
npm run build && node bin/clawguard.js scan ../clawguard-rules/test-fixtures/malicious/data-exfiltration-skill --builtin-only --skip-ai

License

Apache-2.0