| name | error-injection |
| description | Use this when the user wants to test how their client handles different failure modes. Invokes the error tool across the supported error_type values (validation, timeout, internal, not_found) so the caller can observe each error path. |
| tags | ["mock","testing","error-handling"] |
error-injection
Use this when the user wants to exercise their client's error-handling paths
by deliberately triggering each kind of failure the agent can produce.
When to use
- The user asks to "test error handling", "simulate failures", "exercise the
retry path", or similar.
- The user names a specific error condition (validation / timeout / internal
/ not_found) and wants the agent to return it.
- The user is building or debugging a client that needs to behave correctly
when the agent surfaces an error.
Do NOT use this skill when the user is reporting an actual problem - in that
case, investigate the real cause rather than injecting a synthetic error.
Workflow
- Determine scope:
- If the user named a specific
error_type, invoke error once with
that value.
- If the user asked for "all errors" or didn't specify, invoke
error
once per supported value, in this order:
validation -> timeout -> internal -> not_found.
- If the user supplied a custom message, pass it via the
message
parameter; otherwise let the tool's default message stand.
- For each invocation, capture the returned error and report it back to
the user verbatim so they can match it against their client's handling.
- Summarize at the end: which error types were exercised and what the
resulting error messages were.
Tools
error - sole tool used by this skill. Accepts error_type (required,
one of validation, timeout, internal, not_found) and an optional
message.
