The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx @useclawlink/cli, which downloads the latest version of the vendor's command-line interface from the NPM registry at runtime.
[COMMAND_EXECUTION]: All Supabase interactions are performed via shell commands using the ClawLink CLI (npx @useclawlink/cli). This includes logging in, connecting services, and executing database actions.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface through tools that read from external Supabase projects, such as supabase_select_from_table, supabase_run_read_only_query, and supabase_get_project_logs.
Ingestion points: Data returned from Supabase database tables, project logs, and organization details enter the agent's context in SKILL.md via the run command.
Boundary markers: The skill includes an explicit instruction to "Reads first; confirm with the user before any write," establishing a human-in-the-loop boundary for state-changing operations.
Capability inventory: The agent can execute arbitrary shell commands via the CLI, write to Supabase via SQL queries or Edge Function deployment, and list project secrets.
Sanitization: No explicit automated sanitization is described, relying on the user-confirmation guideline and the LLM's internal safety filters.

supabase