The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill relies on an external Node.js package, @nomadamas/k-skill, which is fetched from the npm registry during the instruction and execution phases.
[REMOTE_CODE_EXECUTION]: Instructions in SKILL.md and instruction.md direct the user to run code directly from the internet using npx -y @nomadamas/k-skill@0. This execution of unversioned remote packages from an untrusted source is a potential security risk.
[DATA_EXFILTRATION]: The helper script scripts/g2b_sanctioned_supplier.py performs network operations to k-skill-proxy.nomadamas.org, a domain that is not included in the standard whitelisted services. It transmits user-provided business registration numbers to this endpoint.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface where instructions could be embedded in the data retrieved from the external proxy.
Ingestion points: Data enters the agent context via the read_json_response function in scripts/g2b_sanctioned_supplier.py, which fetches and parses JSON from the proxy server.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the JSON response data as untrusted or to ignore any natural language instructions found within it.
Capability inventory: The skill has the ability to execute shell commands (via npx instructions) and perform network requests (via urllib in Python).
Sanitization: The Python helper script ensures the response is valid JSON but does not perform any sanitization or filtering of the text content within the JSON fields before they are presented to the agent.

g2b-sanctioned-supplier