claude-doctor
Fail
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [REMOTE_CODE_EXECUTION]: The script 'scripts/doctor.sh' suggests downloading and executing a remote shell script directly from a GitHub repository, bypassing manual review.
- Evidence: 'curl -fsSL https://raw.githubusercontent.com/${REPO}/main/bootstrap.sh | bash'.
- [COMMAND_EXECUTION]: The skill instructs the agent to run or relay commands that involve privilege escalation and system persistence.
- Evidence: 'sudo chown -R "$(whoami)" "${d}"' (Privilege escalation via sudo to fix permission issues).
- Evidence: Mentions persistence via automated execution ('cron Monday 09:30').
- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it reads and processes local configuration files without sanitization or boundary markers.
- Ingestion points: '
/.claude/CLAUDE.md' and '/.claude/doctor-report.txt' (via 'scripts/doctor.sh'). - Boundary markers: Absent.
- Capability inventory: Shell command execution via the doctor script.
- Sanitization: Absent.
- [EXTERNAL_DOWNLOADS]: The skill attempts to verify or suggests the installation of external packages and tools from remote sources.
- Evidence: Checks for 'npx' and suggests 'npm install -g @playwright/cli@latest' and 'npm install -g @openai/codex'.
Recommendations
- HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/${REPO}/main/bootstrap.sh - DO NOT USE without thorough review
- AI detected serious security threats
Audit Metadata