lead-generating

Automate lead capture and tracking with Supabase storage and Make.com email workflows, managing conversations from new to qualified status.

This skill appears to do what it says (lead capture, storing conversations, triggering automations) but it asks you to provide a Supabase Service Role Key — a full-access credential that bypasses Supabase row-level security and can read/write any table. Before installing, consider: - Avoid giving the agent the Service Role Key unless the skill runs in a trusted, server-side environment you control. Prefer using an Anon/public key with properly configured RLS policies, or a narrowly scoped server-side token that only allows the required inserts/queries. - Ask the skill author why the Service Role Key is required and whether the skill can work with RLS or a restricted API key. - Verify where the key will be stored by your platform (is it encrypted/secret-scoped?) and whether the agent runtime could leak it to logs or remote endpoints. - Review and/or run the code in a staging environment to confirm the skill only accesses the declared Supabase tables and does not call unexpected external endpoints. There are references to Make.com/Resend in docs, but no hardcoded webhook URLs in the code — you will need to configure those automations yourself. - Rotate the key after testing and monitor Supabase logs for unexpected queries. If you cannot host the skill in a trusted server-side environment or cannot supply a narrowly scoped credential, treat this skill as unsafe to run with a Service Role Key.

Type: OpenClaw Skill Name: sales-bot Version: 1.0.0 The skill is designed for lead management using Supabase and Make.com. It requires a Supabase Service Role Key, which grants full database access, but this is explicitly stated in `SKILL.md` and `skill.json` and is necessary for the skill's stated purpose. The `src/api.ts` code exclusively interacts with the configured Supabase instance for CRUD operations on lead and conversation data, consistently applying `org_id` filters for multi-tenancy. There is no evidence of data exfiltration to unauthorized endpoints, arbitrary command execution, persistence mechanisms, or prompt injection attempts against the agent to perform actions beyond its declared functionality. The `lead.delete` capability is present but is a standard CRUD operation for a lead management system.