The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to execute npx -y @nomadamas/k-skill@0 instruct daangn-cars-search as a mandatory first step. This pattern downloads and runs a package from a public registry at runtime. Furthermore, it delegates the 'primary instructions' for the skill to the output of this command, effectively granting an external source control over the agent's instructional logic.
[EXTERNAL_DOWNLOADS]: The skill relies on external sources for its operational logic and updates, directing the agent to fetch instructions from github.com/NomaDamas/k-skill and use the @nomadamas/k-skill package for execution and maintenance.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection by processing external marketplace data.
Ingestion points: The scripts/daangn_cars.py script fetches vehicle listings, titles, and car descriptions from daangn.com and outputs them to the agent context.
Boundary markers: The skill lacks explicit delimiters or instructions to treat the fetched car listing data as untrusted content.
Capability inventory: The agent environment allows shell command execution (Python, npx) and network access to retrieve data.
Sanitization: There is no evidence of sanitization or filtering of the marketplace content (e.g., stripping markdown or hidden instructions) before it is displayed to the agent.

daangn-cars-search