APIs · apply by default
/api-errors
Standardize useful error responses and propagation
Use to standardize error behavior without changing business policy; api-design defines a new contract.
Make it your own.
In Claude Code, use the slash command and add your context. In Codex, select api-errors from the just-vibe skill picker, then send the same brief.
Version 0.11.0 also supports /jv api-errors, /just-vibe api-errors and /jv:api-errors in Claude. See shortcut setup and context examples.
/just-vibe:api-errors Standardize safe API errors while preserving published machine-readable codes./just-vibe:api-errors Standardize errors while preserving a client's retry behavior on conflict./just-vibe:api-errors Review errors from source without provoking real service failures.What the agent does
- Inventory existing client-visible codes and shapes and the compatibility they must preserve.
- Map domain failures to stable public errors deliberately, redacting internal details while preserving safe correlation IDs.
- Test representative client and server failures: validation, authorization, dependency and unexpected errors.
Inputs
- API scope, current error format, consumers, and logging requirements.
Optional context: scope, references, constraints, successCriteria, environment, mode, budget.
Scope
- Reads
- Error status, machine-readable codes, safe messages, propagation, and trace correlation.
- Writes
- Apply: only the requested local changes and relevant isolated verification. Inspect/plan requests remain inspection/planning. External actions require their exact action and target in session authorization.
- Mode
- Apply; API scope, current error format, consumers, and logging requirements.
- Prerequisites
- Interface definitions, producer/consumer source, authentication model, versioning constraints, and isolated test endpoints. External API calls must respect environment, credentials, rate limits, and side-effect scope.
Expected output
- Error taxonomy and documented contract, the mapping locations, and checks for validation, authorization and dependency failures.
How the work is checked
- Validation errors identify usable field problems; unexpected failures do not leak stack traces or credentials.
When to stop or clarify
- Do not turn all failures into success responses or change public codes silently. Unknown consumer reliance needs compatibility handling.
Handling missing context
- Infer
- Read producer/consumer schemas, error contracts, auth conventions and known supported client versions.
- Assume
- Keep compatible response and pagination semantics where the brief does not request a breaking change.
- Ask
- Ask when contract sources disagree or an unknown consumer changes compatibility; do not require live credentials to write or test an isolated client.
Technical guidance
- Evidence
- Inventory exception sources, status semantics, domain codes, request IDs and retry behavior.
- Method
- Map expected domain failures to stable public errors; redact internal details while retaining correlated server diagnostics.
- Pitfall
- Returning 200 with an error-shaped body or retryable status for a permanent denial misleads clients.
- Check
- Exercise invalid input, forbidden resource, dependency timeout and unexpected exception; verify public redaction and useful internal correlation.
Situational decisions
When changing a code would break a known consumer: Add a compatibility path or a versioned transition instead of silently normalizing it.
When no published error contract exists: Prefer RFC 9457 Problem Details (application/problem+json) with stable type URIs and extension members for field errors; never break an existing published shape.
The coding agent follows this workflow using its available tools. Installation does not grant service access or guarantee an outcome. Read the compatibility notes.