The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to interact with and process data from external Supabase databases and storage buckets, which serves as a potential ingestion point for indirect prompt injection if the stored data contains malicious instructions.
Ingestion points: The agent fetches data from database tables (e.g., invoices, customers) and storage objects as shown in evals/files/edge-function-index.ts and evals/files/storage-usage.ts.
Boundary markers: While the skill emphasizes strict authorization via RLS and JWT verification, it does not explicitly instruct the agent to use boundary markers or delimiters when processing the content of those fields.
Capability inventory: The skill allows the agent to execute supabase CLI commands, run SQL queries, and perform network operations through the Supabase SDK.
Sanitization: The documentation focuses on session security and RLS but does not provide specific guidance on sanitizing the textual content retrieved from the database before it is used in further prompts.
[EXTERNAL_DOWNLOADS]: The skill includes code snippets in the evaluation files that reference external JavaScript libraries from https://esm.sh/. This is a well-known service for serving JavaScript modules, and these references are used within the context of teaching the agent how to audit such code.
[COMMAND_EXECUTION]: The skill defines several legitimate workflows that involve the use of the supabase CLI for project initialization, local stack management (supabase start), database migrations (supabase db push), and edge function deployment.

supabase