The Agent Skills Directory

COMMAND_EXECUTION (HIGH): The script scripts/dev-server.sh is vulnerable to shell command injection. It constructs an ARGS string using unvalidated user input from CLI flags (--host and --port) and then executes it unquoted within the npx medusa develop $ARGS command. An attacker providing a value like 9000; malicious_command could execute arbitrary code on the host system.\n- DATA_EXPOSURE (MEDIUM): Several scripts, including scripts/start-production.sh and scripts/setup-testing.sh, automate the handling of sensitive .env files and environment variables. start-production.sh specifically copies .env to .env.production. Automated secret handling by an agent using these scripts poses a risk of credential exposure if the environment is not strictly isolated.\n- EXTERNAL_DOWNLOADS (LOW): Scripts such as scripts/setup-testing.sh and scripts/start-production.sh perform npm install or use npx to download and execute code from the public npm registry at runtime. While standard for Node.js development, this introduces a dependency on external third-party infrastructure and code.\n- INDIRECT_PROMPT_INJECTION (LOW): Scaffolding scripts (e.g., scripts/create-api-route.sh, scripts/create-module.sh) accept user-provided names for directory and file creation without sanitization. This creates a surface for path traversal where a malicious input could potentially overwrite sensitive files outside the project's intended directory structure.\n
Ingestion points: CLI arguments $1 and $2 in scripts/create-api-route.sh, scripts/create-module.sh, and scripts/create-scheduled-job.sh.\n
Boundary markers: None.\n
Capability inventory: Directory creation (mkdir -p), file writing (cat >), and file modification (sed).\n
Sanitization: None.

medusajs-developer