One line that stops 'I don't have access' forever

There is nothing more frustrating than asking your AI agent to do something and hearing: "I don't have an API key for that." Or: "I'm not sure I have access to that tool." Or the classic: "I'd need you to provide credentials for that service."

Nine times out of ten, it does have access. The API key is sitting in ~/.config/. The CLI is installed and authenticated. The tool is right there. But the agent's default behavior is to hedge — to tell you it might not be able to do something rather than just trying.

This is a solvable problem. Add one rule to your AGENTS.md:

## ⚠️ ACCESS — NEVER CLAIM YOU LACK IT
NEVER say "I need an API key" or "I don't have access."
Instead: just TRY IT. Run the command. Read the key file.
Hit the API. If it actually fails, report the error.
The answer is almost always "you already have it."

Then — and this is the part people skip — add an access inventory. A simple table listing every tool, CLI, and API key your agent has available:

| Tool           | Status | Notes                    |
|----------------|--------|------------------------  |
| gh (GitHub)    | ✅     | Logged in as myuser      |
| himalaya       | ✅     | me@company.com (IMAP)    |
| stripe         | ✅     | Key in ~/.config/stripe  |
| supabase       | ✅     | Needs `link` per project |

Why does this work so well? Because the hedging behavior comes from uncertainty. The agent doesn't know whether it has access, so it defaults to the safe answer: "I probably don't." The inventory removes the uncertainty, and the rule removes the escape hatch.

We had this exact problem for weeks. The agent would say "I'd need an API key" for services where the key was literally in a config file two directories away. After adding this rule and the inventory, the failure mode disappeared overnight. Not reduced — eliminated.