Slow and Steady: I Built Two Tools That Do the Checks I Dread

Description and use

Two open‑source tools, Permit and AgentReady, automate AI agent spending checks and website readability assessments, passing extensive tests and reducing manual oversight.

Problem

Automating repetitive checks for AI agent spending limits and website readability to reduce manual oversight.

Outcome

Permit achieved 406 passed tests on main and 428 passed on hardened branch, blocking over‑limit spends and unapproved orders. AgentReady scored stripe.com 47/100 consistently, with 19 passed tests on a clean clone.

Workflow

• Define spending limits and approval rules in Permit configuration
• Run Permit middleware between AI agent and payment gateway
• Permit blocks over‑limit or unapproved transactions before any money moves
• Use emergency‑stop path to halt running agents
• Configure AgentReady with a target website URL
• AgentReady scans site, evaluates 93 readability checks, outputs a score and fix pack
• Review AgentReady output and apply suggested fixes if needed

Evidence

The full test suite on current main: 406 passed, 3 skipped (including the emergency-stop work merged October 22).

Source

The reference scan — stripe.com — scores 47/100, grade D (44/93 checks).

Source

The agent's ability to voluntarily tighten its own permit existed in the code and passed its tests — but it was never listed in the live agent's instructions.

Source