The Agent Skills Directory

[NO_CODE]: The skill consists entirely of markdown documentation, TypeScript interface definitions, and architectural guidelines. It contains no executable scripts, installers, or remote code triggers.
[PROMPT_INJECTION]: The documentation identifies an indirect prompt injection attack surface where the system processes untrusted data.
Ingestion points: Multiple endpoints ingest external data, including /api/contact (contact form), /api/rag/ingest (document ingestion), and /api/webhooks/paypal (external payment events), as documented in SKILL.md and references/route-inventory.md.
Boundary markers: The provided documentation does not define boundary markers (delimiters) or specific instructions to prevent the agent from obeying instructions embedded within the data processed by these routes.
Capability inventory: The skill describes high-privilege API capabilities including user impersonation (/api/admin/impersonate), system diagnostics (/api/admin/monitoring), and workflow management (/api/admin/n8n), which could be targets for exploitation if an agent is compromised via indirect injection.
Sanitization: Documentation mandates Zod validation for request body structure but does not specify content sanitization or safety filtering for data passed to the AI model.
[COMMAND_EXECUTION]: The route inventory identifies administrative endpoints for remote system management. Specifically, /api/admin/monitoring is documented for "SSH diagnostics, remote actions" and /api/admin/n8n supports actions like "restart" and "rebuild". These findings represent high-risk functionality within the documented API surface but are presented here for governance and architectural visibility.

api-contracts