Outclaw Plan

Build, approve, schedule, monitor, and react to multi-channel B2B outreach campaigns. Reads the user's KB and learned style; constructs multi-step multi-chan...

What to consider before installing: - Missing declarations: The skill expects many external tools and service credentials (Gmail, Slack, WhatsApp, LinkedIn, Calendly, Leadbay) but the registry shows no required env vars or config paths. Ask the author for a full list of required CLIs, environment variables, and filesystem paths and for justification of each. - Silent auto-invocation: The instructions tell the agent to silently auto-invoke other skills (outclaw-research, outclaw-style, shared scripts) without asking the user. If you install this, verify whether those downstream skills are trusted and what they will do (especially any that write to KB or trigger network calls). - Persistent writes and side effects: The skill writes /tmp manifests, updates campaign state, marks drafts as drafted, and logs to Leadbay. Confirm what data will be written, where it will be stored, how long it is retained, and whether you can review/approve those writes before they occur. - Sending and approval: The SKILL.md emphasizes per-touchpoint approval and anti-spam caps, but it also supports a 'trust mode' for pre-approvals. If you do not want automated sending, ensure 'trust mode' is disabled and test that the agent never sends without explicit approval. - Sandbox/testing: Because the skill is instruction-only and relies on host CLIs, test it in a safe sandbox account (test Gmail, test Slack workspace, test Leadbay tenant) before giving access to production accounts. - Ask for source and provenance: The registry homepage links to a GitHub repo; inspect that repo for the referenced scripts (plan_scaffolder.py, validators, schedule_manager.py, channel adapters) to confirm there is no hidden network endpoint or obfuscated behavior. - Least privilege: If you proceed, grant only the minimal credentials needed (dedicated service accounts or test accounts), and audit outgoing network calls and created messages. Require explicit declaration of all required env vars and an install checklist from the publisher. Overall recommendation: do not install in production until the author provides a manifest of required CLIs/credentials, clarifies silent auto-invocation behavior, and shows the scripts that will be executed so you can review persistent writes and external API usage.