Step 1 of 4
Build the error taxonomy
Recovery starts with classification - each class gets a different answer.
Create an error taxonomy for my agent's tools. Classify every known failure into: TRANSIENT (timeouts, rate limits, flaky network - safe to auto-retry), CORRECTIVE (invalid arguments, file not found, lint failure - the model must adjust and retry), FATAL (permission denied at the harness level, missing credentials, unrecoverable environment issues - escalate to the user). For each tool, list its known failure modes and their classes, and mark which operations are non-idempotent (writes, sends, payments) and therefore need idempotency keys or read-back verification before any retry. Deliver the taxonomy table.
Expected after this step
A per-tool error taxonomy with idempotency flags.
Should not happen
- ✕Retrying everything including operations that mutate state twice
- ✕Hiding errors from the model so it reasons from a false world
- ✕No loop breaker: the agent burns its budget on one broken idea
- ✕Silent give-up: the run ends with neither success nor an honest report
Verify before continuing
Do not move on until every check is true. The complete button stays locked until then.
Do not continue if…
- !Retrying everything including operations that mutate state twice
- !Hiding errors from the model so it reasons from a false world
- !No loop breaker: the agent burns its budget on one broken idea
- !Silent give-up: the run ends with neither success nor an honest report
If the AI messes this up
Use this when the AI fakes progress or breaks the feature. It forces a real fix.
Everything is classified transient. Re-check: only failures that can succeed on unchanged retry are transient - a 400 with the same arguments will fail forever and belongs in corrective.