The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses authoritative directives (e.g., "CRITICAL", "knowledge is outdated") to instruct the AI to prioritize the provided content over its internal training. Additionally, it documentation defines a surface for indirect prompt injection through the DurableAgent implementation.
Ingestion points: Untrusted data enters the agent context via the userMessage parameter in the myAgentWorkflow function in SKILL.md.
Boundary markers: The provided code snippets do not demonstrate the use of delimiters or instructions to ignore embedded commands within user input.
Capability inventory: Functions utilizing the "use step" directive are granted "full Node.js access," allowing for significant system capabilities (file access, network ops) if triggered by the agent.
Sanitization: No input validation or sanitization logic is shown in the provided examples.
[COMMAND_EXECUTION]: The skill describes several npx workflow CLI commands for debugging and inspecting workflow runs, which interact with the local environment and Vercel's infrastructure.
[EXTERNAL_DOWNLOADS]: The documentation references external resources from https://useworkflow.dev and https://github.com/vercel/workflow. These are official resources from Vercel, which is a well-known technology provider and trusted organization.

workflow