The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions in SKILL.md direct the agent to run npx -y @nomadamas/k-skill@0, which downloads and executes code from a remote registry. This package has full execution capabilities within the agent's environment.
[COMMAND_EXECUTION]: The skill uses shell commands via npx to fetch instructions, update components, and execute the core Python helper script, allowing for arbitrary command invocation through the CLI tool.
[EXTERNAL_DOWNLOADS]: Dependencies and instructions are retrieved from external sources (NPM and GitHub) at runtime rather than being fully self-contained.
[DATA_EXFILTRATION]: The helper script scripts/nts_business_registration.py sends sensitive data, including business numbers and representative names, to the external domain k-skill-proxy.nomadamas.org for processing.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied business registration data and external API responses, creating a surface for potential indirect prompt injection attacks.
Ingestion points: CLI arguments in scripts/nts_business_registration.py such as --b-no and --business-json (containing p_nm, b_nm, etc.).
Boundary markers: The skill provides guidelines in instruction.md regarding privacy and workflow, but lacks hard technical boundaries between the external data and the agent's prompt context.
Capability inventory: The Python script performs network operations (HTTP POST) and prints raw JSON responses that are interpreted by the agent.
Sanitization: Input validation is performed using regular expressions (e.g., 10-digit number check) and character length limits, providing some protection against basic injection payloads.

nts-business-registration