agent-blueprint/agent-blueprint
Estimate what an AI agent task costs from measured runs, write a one-page summary of what an agent may do and whether it's worth it, and draw an agent's own map of what it can read and change.
Estimate the tokens and dollar cost of an AI agent task or recurring agent job, using measured runs of ten everyday jobs on Claude Code and Codex. Use when someone asks what an agent, agent task or automation will cost to run, how many tokens a task uses, or needs the cost section of an agent business case.
Write a one-page summary of an AI agent for a boss, security reviewer or finance approver, covering what it may do and what enforces each rule, what it costs, and whether it's worth it. Also called an agent card, agent charter or scope document. Use when someone needs to explain, approve or justify an AI agent.
Draw a one-page picture of what this agent can read and change, connection by connection, with who has to approve, its five riskiest powers, what it can't see about itself, and what could be trimmed. Use when the person asks what their agent can access or do, wants to review its permissions, or asks the agent to draw or map itself.