kafka-resilience-and-schema-evolution — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited kafka-resilience-and-schema-evolution (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Generic streaming guidance is not enough for production Kafka. Agents routinely introduce breaking schema changes, under-provisioned durability settings, and missing poison-message isolation. This skill mandates enforceable broker, producer, consumer, and registry guardrails before any production change ships.
Kafka topics, producers, or consumersacks, retries, idempotence)Pair with streaming-and-messaging-systems for broader event design. Pair with avro-protobuf-json-schema-registry when registry subjects and compatibility CI are in scope.
Document:
Require unless explicitly waived with owner approval:
acks=all (or acks=-1)enable.idempotence=true when ordering and deduplication matterretries with delivery.timeout.ms aligned to SLAmax.in.flight.requests.per.connection=1 when strict ordering is requiredBefore any schema change:
BACKWARD, FORWARD, or FULL — not NONE in production)references/kafka-production-guardrails.md for DLQ and evolution patternsEvery production consumer that parses external payloads must define:
Plan for:
Use mcp-data-observability-integration with mcp/kafka.mcp.json to inspect consumer group lag and topic metadata before changing consumer code or partition counts.
| Rationalization | Reality |
|---|---|
| "acks=1 is fine because Kafka is durable." | Leader acknowledgment without full ISR acknowledgment loses events under failure scenarios. |
| "We can fix schema breaks by redeploying consumers quickly." | Breaking changes propagate to many consumers and batch sinks before redeploy completes. |
| "DLQs add too much operational overhead." | Poison pills without DLQs stall partitions, inflate lag, and hide data loss as consumer retries. |
| "Schema compatibility NONE is okay for internal topics." | Internal topics still feed warehouses, stream processors, and audit systems. |
NONE compatibilityacks=0 or acks=1 without documented waiveracks=all and idempotence requirements~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.