daiso-product-search

Warn

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs the agent to execute remote code using npx to retrieve primary instructions and file lists from a vendor-controlled package (@nomadamas/k-skill). Evidence found in SKILL.md.
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the daiso-product-search npm package from a public registry as documented in instruction.md.
  • [CREDENTIALS_UNSAFE]: Contains a hardcoded encryption key (PRE_AUTH_ENC_KEY) used for generating authentication tokens for the Daiso Mall API, as specified in the stock check workflow in instruction.md.
  • [COMMAND_EXECUTION]: The skill relies on shell command execution for package management and fetching dynamic instructions via npx.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided store names and product keywords to perform API queries without explicit boundary markers or sanitization logic. 1. Ingestion points: User inputs for storeQuery and productQuery defined in instruction.md. 2. Boundary markers: Absent; no instructions are provided to delimit or treat user input as untrusted. 3. Capability inventory: Network operations (daisomall.co.kr API) and shell command execution via npm/npx across SKILL.md and instruction.md. 4. Sanitization: Absent; user queries are passed directly to API-calling functions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 17, 2026, 06:35 AM
Security Audit — agent-trust-hub — daiso-product-search