react-doctor
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill fetches a remote markdown "playbook" from https://www.react.doctor/prompts/react-doctor-agent.md and explicitly directs the agent to "follow every step in it." This functionally executes remote instructions retrieved at runtime.
- [REMOTE_CODE_EXECUTION]: Fetches per-rule fix recipes on demand from https://www.react.doctor/prompts/rules//.md to guide the agent's code modification tasks.
- [COMMAND_EXECUTION]: Executes shell commands via npx react-doctor@latest, which downloads and runs the latest version of the vendor's package from the NPM registry.
- [COMMAND_EXECUTION]: Uses curl to retrieve remote instruction files from the vendor's domain, react.doctor.
- [INDIRECT_PROMPT_INJECTION]: The agent is designed to ingest and obey instructions fetched from external URLs, creating a vulnerability surface for indirect injection if the remote source is compromised.
- Ingestion points: Remote fetches in SKILL.md for react-doctor-agent.md and per-rule markdown files.
- Boundary markers: None identified in the instructions for following the remote playbook.
- Capability inventory: Shell execution (npx, curl), file system modification (working tree edits), and browser interaction (CDP).
- Sanitization: No sanitization or validation of the remote instruction content is performed before the agent is told to follow it.
- [EXTERNAL_DOWNLOADS]: Fetches configuration and instructions from the vendor's official domain (react.doctor) and the public NPM registry.
Audit Metadata