The Agent Skills Directory

[COMMAND_EXECUTION]: The skill describes using the prefect CLI for workflow mutations, such as triggering deployments (prefect deployment run) and canceling flow runs (prefect flow-run cancel). It recommends safety measures including verifying the client's execution capability, checking authentication status, and requiring explicit user requests before execution.- [PROMPT_INJECTION]: The skill ingests untrusted data from multiple sources including event streams, flow run logs, and documentation search tools. This creates an indirect prompt injection surface where adversarial content in external data could influence agent actions. Evidence chain: 1. Ingestion points: MCP tool outputs (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Shell command execution via the prefect CLI. 4. Sanitization: Absent.

workflows