skills/socketdev/sauce/socket-inspect/Gen Agent Trust Hub

socket-inspect

Fail

Audited by Gen Agent Trust Hub on Oct 14, 2026

Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill contains a hardcoded functional API token (sktsec_t_--RAN5U4ivauy4w37-6aoKyYPDt5ZbaT5JBVMqiwKo_api) for the Socket public demo. Hardcoding credentials in instruction files is a dangerous practice that exposes functional secrets and can lead to unauthorized usage tracking or credential misuse.
  • [COMMAND_EXECUTION]: The skill directs the agent to execute local shell commands using pnpm dlx socket and curl. This gives the agent the capability to interact with the underlying system and execute external binaries.
  • [EXTERNAL_DOWNLOADS]: The skill uses pnpm dlx to dynamically download and execute the socket package from the NPM registry. It also makes network requests to api.socket.dev and browse socket.dev using WebFetch. These resources are owned by the vendor and are documented neutrally.
  • [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted package metadata (names, versions, and ecosystems) and interpolates them directly into shell commands and API requests.
  • Ingestion points: Package identifiers (ecosystem, name, version) supplied by users or extracted from external files (SKILL.md).
  • Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat package metadata as untrusted or to ignore instructions embedded within that metadata.
  • Capability inventory: The skill possesses high-privilege capabilities including shell execution (pnpm dlx), web requests (curl), and automated browsing (WebFetch).
  • Sanitization: The skill lacks explicit instructions for sanitizing or validating user-provided strings before they are executed in the shell environment, which could allow malicious package names to execute arbitrary commands.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Oct 14, 2026, 02:51 AM