cliproxy-deploy

Warn

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONCREDENTIALS_UNSAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: Downloads source code from an unverified GitHub repository (github.com/router-for-me/CLIProxyAPI) and executes it using the go run command.
  • Evidence: git clone https://github.com/router-for-me/CLIProxyAPI.git /root/CLIProxyAPI && go run ./cmd/server in SKILL.md.
  • [EXTERNAL_DOWNLOADS]: Fetches executable content and toolchains from external sources during the deployment process.
  • Evidence: Clones the repository from github.com/router-for-me/CLIProxyAPI and triggers automatic Go toolchain downloads via GOTOOLCHAIN during code execution.
  • [COMMAND_EXECUTION]: Performs several sensitive system-level operations on a remote host via SSH.
  • Evidence: Modifies firewall rules using ufw allow <PORT>/tcp.
  • Evidence: Installs system packages using apt-get install -y golang-go git.
  • Evidence: Terminates specific processes using pkill -f 'cmd/server -config'.
  • Evidence: Persistently modifies user shell profiles (~/.zshrc or ~/.bashrc) to store environment variables.
  • [CREDENTIALS_UNSAFE]: Handles and transfers sensitive authentication materials between local and remote environments.
  • Evidence: Syncs OAuth credential files (~/.cli-proxy-api/*.json) from the local machine to a remote VPS via scp.
  • Evidence: Manages SSH private keys used for remote server access.
  • Evidence: Stores generated API keys in plaintext within config.yaml and shell environment variables.
  • [DATA_EXFILTRATION]: The core functionality involves moving local account credentials (OAuth tokens for third-party services) to an external server to be exposed as an API endpoint.
  • [PROMPT_INJECTION]: Vulnerable to indirect prompt injection (Category 8) due to the lack of sanitization when processing data from the deployed service.
  • Ingestion points: Reading output from the deployed API via curl ... /v1/models | head -c 400 in SKILL.md.
  • Boundary markers: Absent.
  • Capability inventory: The skill has extensive system capabilities including Bash tool access, SSH command execution, package installation, and file system writes.
  • Sanitization: No evidence of validation or escaping for the ingested content.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 2, 2026, 11:20 AM
Security Audit — agent-trust-hub — cliproxy-deploy