The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches the Socket CLI and cdxgen from the npm registry at runtime using npx (e.g., npx socket, npx @cyclonedx/cdxgen). This ensures the latest version is used but involves downloading and executing remote code.
[COMMAND_EXECUTION]: The skill executes multiple shell commands to perform dependency scans and audits. This includes running npx socket, npx @cyclonedx/cdxgen, and native audit tools such as npm audit, pnpm audit, and yarn audit.
[CREDENTIALS_UNSAFE]: A hardcoded public demo API token (sktsec_t_--RAN5U4ivauy4w37-6aoKyYPDt5ZbaT5JBVMqiwKo_api) is provided in the instructions and setup steps. While this is a vendor-supplied demo token for unauthenticated access, hardcoding credentials in skill files is generally discouraged.
[DATA_EXFILTRATION]: Project dependency information is uploaded to the Socket vendor dashboard (socket.dev) when the skill is run in persistent mode (npx socket scan create .). Temporary mode (--tmp) returns results locally, but the persistent mode is a documented feature for authenticated users.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes untrusted project manifest files (e.g., package.json, requirements.txt) using external tools.
Ingestion points: Project manifest and lock files (automatically detected by socket and cdxgen in the target directory).
Boundary markers: None detected; the skill does not use specific delimiters to separate untrusted file content from instructions.
Capability inventory: The skill can execute shell commands via npx and read/write files (e.g., creating bom.json).
Sanitization: No explicit sanitization or validation of the manifest file contents is performed before processing.

socket-scan