The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill delegates its primary functionality—including interview logic, document rendering, and file installation—to scripts fetched from a third-party GitHub repository (kimlawtech/korean-privacy-terms). While the skill uses a pinned commit SHA in scripts/upstream.pin to manage integrity, the actual logic is not contained within the skill and is executed at runtime.
[EXTERNAL_DOWNLOADS]: The scripts/install.sh script is designed to clone an external repository from GitHub (https://github.com/kimlawtech/korean-privacy-terms.git) and copy its contents into the agent's environment (e.g., ~/.claude/skills/).
[INDIRECT_PROMPT_INJECTION]: The skill facilitates an interview process that ingests user-supplied information to generate legal documents (Next.js components and MDX files). This pattern presents a surface for indirect prompt injection if the user-provided data is not properly sanitized before being interpolated into the generated source code templates.
Ingestion points: User input collected during the mandatory interview steps (e.g., service type, company name, representative info).
Boundary markers: The skill mentions inserting legal disclaimers, but no specific prompt boundary markers are defined for the interpolation process.
Capability inventory: File system modification (mkdir, rm, rsync, cp), network operations (git clone), and command execution (npx).
Sanitization: None explicitly implemented in the skill; responsibility is delegated to the external upstream logic.
[COMMAND_EXECUTION]: The skill instructions repeatedly use npx -y @nomadamas/k-skill@0 to fetch instructions and execute scripts. Although the package is associated with the skill's author (nomadamas), this involves running remote code from the NPM registry to determine skill behavior.

korean-privacy-terms