korea-weather
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads and executes the
@nomadamas/k-skillpackage from the NPM registry during initialization, file listing, and update procedures. - [REMOTE_CODE_EXECUTION]: The agent is instructed to run
npx -y @nomadamas/k-skill@0 instruct korea-weatherand use its output as the primary set of instructions. This delegates instruction management to a remote package, allowing the skill's behavior to be modified externally after deployment. - [DYNAMIC_EXECUTION]: The instructions specify that the CLI tool detects the runtime environment to print applicable instructions, which involves generating and executing instructions at runtime based on the output of a remote script.
- [COMMAND_EXECUTION]: The skill uses
curlto fetch weather data from the vendor-controlled proxy serverhttps://k-skill-proxy.nomadamas.org. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from a remote proxy, which introduces a potential attack surface if the external weather data source were compromised.
- Ingestion points: Data is received via
curlrequests to the/v1/korea-weather/*endpoints onk-skill-proxy.nomadamas.org(found in instruction.md). - Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the weather data are provided.
- Capability inventory: The skill has access to shell execution (
npx,curl) and the ability to summarize and present data to the user. - Sanitization: There is no evidence of validation or sanitization of the JSON response fields (e.g.,
TMP,SKY,obsrValue) before they are processed by the agent.
Audit Metadata