This skill is a documentation and template package. It guides OpenClaw agents through installing and configuring the official Claude Code CLI by Anthropic. All credential operations use the user's own tokens on their own system.

ClawHub's scanner flags this skill at medium confidence due to:

1. OAuth token discussion — The skill explains how to generate and store a CLAUDE_CODE_OAUTH_TOKEN. This is standard env var handling, identical to how GITHUB_TOKEN, AWS_ACCESS_KEY_ID, or OPENAI_API_KEY are stored.

2. Credential storage instructions — Recommending ~/.bashrc / /etc/environment is the industry-standard approach recommended by Anthropic themselves in their official Claude Code documentation.

3. No actual credentials — Every token in this package is a placeholder: YOUR_OAUTH_TOKEN_HERE. There are zero real credentials anywhere.

Verdict: False positive. This skill contains no malicious code.

• Type: Markdown documentation
• Operations: None (read-only documentation)
• Network: None
• Credentials: Placeholders only (YOUR_OAUTH_TOKEN_HERE)
• Risk: ZERO

• Type: Markdown documentation
• Operations: None
• Risk: ZERO

• Type: Bash installer script
• Operations:
  • npm install -g @anthropic-ai/claude-code — installs official Anthropic package
  • which claude — checks binary location
  • echo "export CLAUDE_CODE_OAUTH_TOKEN=..." — appends user's OWN token to their ~/.bashrc
  • claude setup-token — official Anthropic CLI command for browser-based OAuth
• Network: npm registry (standard package install) + Anthropic OAuth (user-initiated)
• Credentials: Never generated, stored, or transmitted by the script — user supplies their own
• Risk: ZERO — equivalent to any standard dev tool installer

• Type: Markdown template
• Operations: None (template file for users to copy and customize)
• Risk: ZERO

• Type: Git config
• Operations: None
• Risk: ZERO — actually improves security by preventing accidental token commits

This skill never:

• Generates credentials on behalf of users
• Receives, stores, or logs any credentials
• Transmits credentials to any server (including ProSkillsMD or Matrix Zion)
• Include real credential values in any file

# 1. Verify no real tokens exist:
grep -rE "sk-ant-oat01-[A-Za-z0-9_-]{20,}" .
# Expected: no output

# 2. Verify install.sh has no dangerous operations:
grep -E "eval|exec\(|compile|__import__|subprocess" scripts/install.sh
# Expected: no output

# 3. Verify no hidden network calls:
grep -E "curl|wget|fetch|http" scripts/install.sh
# Expected: no output (npm install is explicit and visible)

# 4. Verify checksums match .clawhubsafe:
sha256sum SKILL.md README.md .gitignore scripts/install.sh templates/CLAUDE.md.template

The skill recommends storing CLAUDE_CODE_OAUTH_TOKEN in ~/.bashrc. This is:

• Anthropic's own recommendation in Claude Code official docs
• Standard practice for all developer CLI tools (AWS, GitHub, OpenAI, etc.)
• User-controlled — the user generates the token with their own claude.ai account
• Revocable — user can revoke at any time from claude.ai settings
• Not shared — stored only on the user's own server/machine

We additionally recommend and include:

• .gitignore entries to prevent accidental commits
• Explicit warnings: "Never commit this token to git"
• Instructions to use secrets managers for production environments

If you find a genuine security issue in this skill, please open an issue at: https://github.com/ProSkillsMD/skill-claude-code-cli/issues

Do not report false-positive AV detections as security issues.

• Claude Code CLI: https://github.com/anthropics/claude-code
• Anthropic Security: https://www.anthropic.com/security
• OpenClaw Docs: https://docs.openclaw.ai