add-iron-proxy
Fail
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The script
scripts/install-claude.shdownloads and executes a setup script fromhttps://claude.ai/install.shviacurl | bash. This operation is used to install the Claude CLI.\n- [EXTERNAL_DOWNLOADS]: The skill fetches source code from theironsh/iron-proxyGitHub repository to build the gateway locally. It also pulls container images for PostgreSQL and the Iron Control console. These resources are pinned by commit hashes or digests.\n- [INDIRECT_PROMPT_INJECTION]: The gRPC approval bridge inpayload/src/gateway-providers/iron-proxy-approval.tsparses metadata from outbound requests to populate approval cards. This constitutes a surface where untrusted external data is processed and presented to the operator.\n - Ingestion points: gRPC metadata key
x-iron-approval-summaryinIronProxyApprovalBridge.#transformRequest.\n - Boundary markers: Present; the skill uses structured JSON summaries and specific presentation logic for decisions.\n
- Capability inventory: The skill is capable of Docker builds, image management, and intercepting network traffic.\n
- Sanitization: Implements
safeDisplayto filter character input and validates JSON schemas for summaries.\n- [COMMAND_EXECUTION]: The setup process uses the Docker CLI for orchestration, including building images from source, running containers with restricted capabilities, and managing persistent data volumes.
Recommendations
- HIGH: Downloads and executes remote code from: https://claude.ai/install.sh - DO NOT USE without thorough review
Audit Metadata