The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local commands using child_process utilities to manage its environment, project state, and workflow deployments.
It invokes the vercel, git, and gh CLIs to handle project configuration, repository operations, and Vercel deployments as part of its core automation logic.
It executes the claude and codex CLIs to run agent stages using local user subscriptions when configured for local execution.
It manages a dedicated local Postgres database by spawning initdb and pg_ctl binaries bundled in the embedded-postgres package.
[EXTERNAL_DOWNLOADS]: The skill interacts with external services to perform data enrichment and content retrieval.
It communicates with the api.monid.ai and api.blitz-api.ai endpoints to fetch enriched person and company data.
It uses a custom fetching utility to retrieve readable text from public web addresses provided as input to workflows.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which constitutes an indirect prompt injection surface.
Ingestion points: External data enters the agent context through user-uploaded CSV files and inbound webhooks for workflow triggers.
Boundary markers: The skill uses standard AI SDK messaging structures and tool-calling protocols to separate instructions from data.
Capability inventory: Workflows have the capability to perform network operations, write to database tables, and spawn local subprocesses.
Sanitization: The skill implements sanitization routines to remove NUL characters from database inputs and uses an SSRF-resistant fetcher that validates hostnames against private and internal IP ranges before execution.

gtm-workflow