The Agent Skills Directory

[SAFE]: The skill is focused on auditing subscription mechanics such as trial conversions, renewal disclosures, and cancellation paths to ensure honesty and technical consistency.
[EXTERNAL_DOWNLOADS]: The skill references a directory of official government portals and international standards organizations (e.g., FTC, EUR-Lex, W3C, NIST, and UAE government portals) to support compliance research. These sources are well-known, authoritative, and trusted.
[INDIRECT_PROMPT_INJECTION]: The skill processes external content like website copy, checkout flows, and terms of service, which constitutes a potential attack surface.
Ingestion points: Untrusted data enters the agent context through the inspection of billing pages, legal terms, and transactional emails (documented in SKILL.md).
Boundary markers: The skill mandates the use of evidence labels (e.g., OBSERVED, DECLARED, INFERRED) to help the agent distinguish between observed technical facts and declared external claims.
Capability inventory: The skill interacts with billing provider APIs (test mode clocks) and retrieves content from official regulatory websites.
Sanitization: The instructions require citing specific sources with access dates and keeping quotes short to prevent external data from overriding agent instructions.

subscription-readiness