Triggers
Trigger proposal validation errors
Correct common structural and scoped-reference problems in a v1 trigger proposal.
These issue codes define the v1 proposal-validation vocabulary. They help an AI or human revise a planning document. Gooshi does not currently expose this validator as a public API, and the current trigger editor may use different wording.
| Code | Meaning | Correction |
|---|---|---|
invalid_contract | A required field, value, type, or version does not match the strict schema. | Compare the proposal with the v1 JSON Schema and remove unknown fields. |
forbidden_authority_field | The proposal tries to select publication, activation, environment, account, or routing authority. | Remove the field. A proposal can describe intent only. |
unknown_tool_ref | A tool reference is not available in the current scope. | Refresh discovery and use only a reference from Gooshi's future scoped export; do not invent one. |
unknown_field_ref | A comparison uses a field reference not present in the supplied tool context. | Leave the field unresolved until Gooshi supplies it through the scoped export. |
unknown_meter_ref | A quota or usage block uses an unavailable meter. | Leave the meter unresolved until Gooshi supplies it through the scoped export. |
unsupported_currency | The currency is not available for the current setup. | Select a supported currency shown in Gooshi. |
amount_out_of_range | The amount is outside the allowed bounds. | Choose an amount within the bounds shown during configuration. |
quota_usage_required | A quota has no counting behavior. | Add usage and choose on_success or on_attempt. |
input_too_deep | The proposal contains excessive nesting. | Simplify nested all and any groups. |
input_too_large | The proposal contains too many values. | Split the business rule into smaller proposals for review. |
cyclic_input | The submitted value is not plain JSON data. | Serialize it as ordinary JSON without object cycles. |
Revise and test
After correcting structural errors, still verify scope and behavior:
- Confirm that every reference came from the current workspace.
- Test a non-matching case.
- Test the first matching case.
- Test values on both sides of every threshold.
- For quotas, test immediately before and after the allowance.
- Review how failed tool calls affect usage.
Schema validity does not prove the business rule is correct. An AI should recommend test cases, but it must never claim a proposal was validated, simulated, saved, activated, or published unless Gooshi visibly reports that result.