The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill communicates with https://slack.nanoclaw.dev to provision Slack applications and generate agent avatars. This is a vendor-owned service ('nanocoai') central to the skill's functionality. Additionally, the installation process fetches companion files from the repository's channels branch.
[COMMAND_EXECUTION]: The skill uses local shell commands during installation (e.g., pnpm run build) and service management. The slack-agent-flow-finish.ts script executes setup/lib/restart.sh to apply new configuration by restarting the host process.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user or agent-supplied data such as room names and purposes, which are interpolated into Slack canvas contracts and system nudges.
Ingestion points: Arguments for the create_agent and create_room tools, specifically the name and purpose fields.
Boundary markers: The room purpose is wrapped in blockquote markers (> ) within the generated Slack canvas markdown.
Capability inventory: The skill possesses capabilities to write to the .env file, modify the central database, and perform network requests to Slack and the provisioning broker.
Sanitization: Input is sanitized via normalizeName for identifiers, and the publicPurpose helper truncates purpose strings to 80 characters. Most importantly, all agent-initiated room and creation actions are protected by an admin-approval guard that prevents unauthorized execution of these capabilities.

slack-agent-flow