How the error reaches the agent
When a tool cannot complete its action, IaGenify records the failed call and passes the error text into the conversation. This gives the model a chance to understand whether the issue is a missing file, invalid input, unavailable service, or a permission refusal. The error is evidence for the next decision, not a successful result.
The agent may then choose a different tool, adjust its arguments, ask you for missing information, or stop and report that it could not proceed. Whether it can recover depends on what failed and which tools are available. A permission refusal is a boundary to respect, not an invitation to repeat the same action until it succeeds.
Review the result before relying on it
A later answer may still be incomplete even if the run continues after an error. Check the final report for the failed operation, any fallback used, and whether the requested outcome was actually confirmed. If the agent says it completed a change, look for a tool result or other evidence that verifies the change rather than inferring success from its intention.
For a useful recovery, give the agent the exact error and the relevant context, such as the file path, service, or operation you expected. Avoid sharing credentials or private tokens. If the same error persists, correct its cause or choose a supported path instead of repeatedly retrying an unchanged request.