The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by reading untrusted data from website-project.json and interpolating it directly into markdown templates.
Ingestion points: Data is read from website-project.json (fields such as legal_config.legal_entity.legal_name, company.name, etc.) and used to fill {{placeholders}} in legal templates.
Boundary markers: No explicit boundary markers or delimiters are used during interpolation to separate data from instructions.
Capability inventory: The skill possesses Write, Edit, and Bash tools, and the generated markdown is explicitly intended for processing by a downstream "page-generator" agent.
Sanitization: The skill performs plain string replacement and does not describe any sanitization, escaping, or validation of the input data before it is written to the legal markdown files.

website-legal