setup-codemap-cli

================================================================================

🔴 VERDICT: HIGH

This skill presents a HIGH security risk primarily due to its instruction to configure the agent to automatically execute commands at the start of every session, and its reliance on installing software from unverified external sources.

Total Findings: 3

🔴 HIGH Findings: • Persistence Mechanism / Command Execution

• Line 116: The skill instructs the user to modify the agent's settings file (~/.claude/settings.json or project-level equivalents) to include a session-start hook. This hook executes commands (codemap hook session-start && echo 'git diff:' && git status) automatically at the beginning of each session. While the current commands are benign, this mechanism allows for arbitrary command execution to be persisted within the agent's configuration, posing a significant risk for future malicious activity.

🟡 MEDIUM Findings: • Unverifiable Dependencies / Command Execution

• Line 48: The skill instructs to install codemap via Homebrew using brew tap JordanCoin/tap && brew install codemap. The JordanCoin GitHub user/organization is not on the list of trusted sources. This means the skill is instructing the agent to download and execute code from an unverified external source, which could contain malicious payloads. • Unverifiable Dependencies / Command Execution
• Line 53: The skill instructs to install codemap via Scoop using scoop bucket add codemap https://github.com/JordanCoin/scoop-codemap and scoop install codemap. Similar to the Homebrew installation, JordanCoin is not a trusted source, introducing a risk of executing unverified external code.

🔵 LOW Findings: • No LOW findings.

ℹ️ TRUSTED SOURCE References: • External Documentation Download

• Line 40: The skill instructs to load documentation from https://raw.githubusercontent.com/JordanCoin/codemap/refs/heads/main/README.md. While the content is from a non-trusted GitHub user (JordanCoin), the domain raw.githubusercontent.com is generally whitelisted for content downloads. This specific action is considered low risk as it only involves reading a README file, but the source is noted.

================================================================================