# Model Usage

Scanned on 2/7/2026

⚠️62/100medium

Script executes external commands and reads files with minimal input validation, presenting moderate security risks through command injection and path traversal vulnerabilities.

## Security Analysis

This script has moderate security concerns primarily around command execution and file system access. While it avoids shell injection by using shell=False and doesn't make external network calls, it has three main vulnerabilities: (1) The provider argument is passed to subprocess without validation beyond argparse choices, which could be bypassed if called programmatically; (2) The input_path parameter allows reading arbitrary files without path validation, enabling potential information disclosure; (3) No resource limits on subprocess output or stdin reading could lead to DoS. The script shows some security awareness with type checking and proper encoding, but lacks defense-in-depth measures. The code is incomplete (cuts off mid-argument definition), preventing full analysis of all input vectors. For production use, implement strict input validation, path sanitization, and resource limits.

## Security Findings

high - code-execution

Subprocess execution with user-controlled provider argument without proper validation

Evidence:

`cmd = ["codexbar", "cost", "--format", "json", "--provider", provider]
subprocess.check_output(cmd, text=True)`

Recommendation: Implement strict allowlist validation for provider argument before passing to subprocess. Use shell=False (already done) and validate provider against hardcoded list of allowed values before subprocess call.

medium - filesystem

Arbitrary file read through input_path parameter without path validation

Evidence:

`with open(input_path, "r", encoding="utf-8") as handle:
    raw = handle.read()`

Recommendation: Validate input_path to ensure it's within expected directories, resolve symlinks, and check for path traversal attempts (../, absolute paths). Consider using pathlib.Path.resolve() and checking against allowed base directories.

medium - code-execution

JSON parsing of external command output without size limits

Evidence:

`output = subprocess.check_output(cmd, text=True)
payload = json.loads(output)`

Recommendation: Add timeout parameter to subprocess.check_output() and implement size limits on output before parsing to prevent resource exhaustion attacks.

low - filesystem

Unrestricted stdin reading without size limits

Evidence:

`if input_path == "-":
    raw = sys.stdin.read()`

Recommendation: Implement maximum size limit when reading from stdin to prevent memory exhaustion: sys.stdin.read(MAX_SIZE) with appropriate error handling.

info - other

Limited error context in exception handling may hide security issues

Evidence:

`except Exception:
    return None`

Recommendation: Log exceptions or provide more specific exception handling to aid in security monitoring and debugging.

## ✅ Good Security Practices

- ✓ Uses shell=False in subprocess calls, preventing shell injection
- ✓ Implements type checking on parsed JSON data structures
- ✓ No use of eval() or exec() functions
- ✓ No hardcoded credentials or API keys
- ✓ No network calls to external endpoints
- ✓ Uses text mode for subprocess output instead of binary
- ✓ Proper encoding specified for file operations (utf-8)

## Source Information

ClawHub Page: [clawhub.com/skills/model-usage](https://clawhub.com/skills/model-usage)

Source URL: [/opt/homebrew/lib/node_modules/openclaw/skills/model-usage](/content/opt/homebrew/lib/node_modules/openclaw/skills/model-usage/index.html)

Code Size: 10,517 characters

Semgrep Findings: 0