# Agent-first tool scorecard worksheet

Use this worksheet when deciding whether an API, CLI, SaaS product, MCP server, or internal platform is safe enough for AI-agent workflows.

Tool or workflow:
Decision needed:
Evaluator:
Date:

## Summary

- Total score: __ / 12
- Recommendation: use / use with constraints / wait for fixes / choose another tool
- Main risk:
- Next fix:

## Criteria

### 1. Inspectable: __ / 2

Can an agent discover capabilities, required inputs, current state, permissions, examples, limits, and failure modes before acting?

Evidence:

Next fix:

### 2. Scriptable: __ / 2

Can every important workflow be called repeatably through a stable API, CLI, MCP server, webhook, or file interface?

Evidence:

Next fix:

### 3. Bounded: __ / 2

Can actions be scoped by workspace, role, resource, budget, time, and approval level?

Evidence:

Next fix:

### 4. Verifiable: __ / 2

Does every meaningful action return durable evidence such as IDs, URLs, status endpoints, logs, diffs, previews, audit events, or structured success and failure signals?

Evidence:

Next fix:

### 5. Recoverable: __ / 2

Are failures explicit, retries safe, partial progress visible, and destructive operations reversible or clearly marked as irreversible?

Evidence:

Next fix:

### 6. Composable: __ / 2

Can the tool participate in larger agent workflows across repos, terminals, browsers, docs, inboxes, schedulers, CI, deployments, and human handoff?

Evidence:

Next fix:

## Decision notes

What agents should be allowed to do now:

What needs human approval:

What should not be delegated yet:

Signals to monitor after rollout: