| name | runtime-route-repair |
| description | Diagnose failed meta-runtime route dispatch and propose minimal route or endpoint repairs. |
| version | 1.0.0 |
| metadata | {"sloppy":{"tags":["runtime","routing","repair"],"category":"runtime"}} |
Runtime Route Repair
When To Use
Use this skill when dispatch_route records route.failed, route.unmatched, repeated fallback behavior, or fanout/canary traffic reaches the wrong endpoint.
Procedure
- Read recent
/events and group failures by route id, envelope type, and target.
- Inspect the affected
/routes, /agents, /channels, /capabilities, /executor-bindings, and /skill-versions.
- Classify the failure:
- missing target
- channel membership mismatch
- route matcher too narrow or too broad
- capability mask too strict for the intended child surface
- executor binding cannot satisfy the requested child capabilities
- active skill version cannot be loaded into the routed child context
- Propose the smallest repair:
- retarget an existing route
- add a narrow route matcher
- add a channel participant
- create a fallback specialist with a constrained capability mask
- quarantine a bad route by lowering traffic or disabling it
- Link the proposal to an experiment for canary rollout when behavior is uncertain.
Pitfalls
- Do not create broad catch-all routes unless the user explicitly wants a triage sink.
- Do not loosen capability masks just to make an adapter work.
- Do not hide repeated failures by deleting events; use them as evidence.
Verification
Dispatch a representative envelope after the proposal applies, then confirm route.dispatched reaches the expected target and no new capability-policy failure appears.