anlyon-troubleshooting

The Agent Skill for diagnosing a failing Anlyon call from its error code or invocation result, with the fix for each, and without ever printing a key.

anlyon-troubleshooting teaches your coding agent to read an Anlyon error by its code, find the cause, and fix it, or tell you exactly what to change in the console when the fix is yours to make.

What it teaches

• How to read an error: match on error.code, keep the requestId for support, and read the body of a 200, where a refused call reports itself.
• What not to do while debugging: print a key to check it, retry an unknown, or raise a limit, disable a policy or ungate an action to make a call pass.
• A short workflow: get the code, look it up, check the key’s identity with auth.identity(), apply the fix, re-run the one call.
• The usual suspects at a glance, from AUTHENTICATION_ERROR to END_USER_NOT_CONNECTED and impact_limit_exceeded.
• MCP and connect-link problems: missing tools, sign-in in CI, used-up links, a missing OAuth app.

When it loads

When an Anlyon call returns an error code such as AUTHORIZATION_ERROR, NOT_FOUND, PAYMENT_REQUIRED, END_USER_REQUIRED, END_USER_MISMATCH, REDIRECT_URL_REQUIRED, PREVIEW_MISMATCH or idempotency_key_reuse, when an invocation is refused or stuck in unknown, or when the MCP server or a connect link misbehaves.

What’s inside

Download: anlyon-troubleshooting.zip.

Install

npx skills add https://docs.anlyon.com --skill anlyon-troubleshooting

The same codes are documented for people in Error handling.

Try it

The invoke in jobs/sync.ts fails with 403. Find out why.

Every call from staging returns NOT_FOUND for an action I can see in the console.

An invocation has been unknown since last night. What do we do?