custom-kafka-consumer

Set up, configure, run, and debug a custom Kafka consumer that streams data into Snowflake via Snowpipe Streaming SDK v2. Use when: building a Kafka-to-Snowflake streaming pipeline, running the CDR demo, or troubleshooting the custom consumer. Triggers: kafka consumer, kafka snowflake, snowpipe streaming kafka, CDR demo, kafka to snowflake, custom consumer, streaming ingest kafka.

2.32x

Quality

82%

Does it follow best practices?

Impact

100%

2.32x

Average score across 2 eval scenarios

Securityby

Passed

No findings from the security scan

Quality

Content

77%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is an exceptionally actionable, well-sequenced runbook with genuine validation gates and teardown safety — workflow clarity and actionability are close to ideal. Its weaknesses are structural: at ~28KB monolithic, with inlined reference material and triple-repeated forecast recipes, it spends tokens it does not need to and ignores progressive disclosure entirely (no references/ bundle at all).

Suggestions

Move the 'Key Consumer Patterns (Reference)' section (client init, channel open/offset recovery, retry table, health check) and the full semantic-view DDL into references/ files (e.g. references/consumer-patterns.md, references/semantic-view.sql), keeping only a short pointer plus key gotchas in SKILL.md.

Collapse steps 7d and 7e into one templated recipe for 'any per-tower daily metric' (view → SNOWFLAKE.ML.FORECAST → 7-day prediction → inspect), showing the drop-rate case in full and the call-volume/data-usage cases as parameter substitutions — this alone would remove ~80 duplicated lines.

Trim per-column COMMENT clauses and the redundant USE DATABASE/USE SCHEMA pairs repeated before each FORECAST call; state the USE convention once.

Dimension	Reasoning	Score
Conciseness	Mostly executable content with little concept-explanation padding, but it could be tightened significantly: sections 7d and 7e repeat the identical view/train/predict/inspect SQL pattern from 7a/7b with only column names changed, and the ~120-line semantic-view DDL with per-column COMMENTs is bulk Claude does not need inline. Not a 2 because almost every line is demo-specific executable material, not filler.	3 / 5
Actionability	Everything is copy-paste ready: exact brew/kafka-topics/mvn commands, complete SQL for table/view/FORECAST/semantic-view creation, full properties files, an interactive producer command table, expected outputs with concrete numbers, and a teardown script. Verification queries (row-count gate, IFF status) cover the common cases.	5 / 5
Workflow Clarity	Steps 1–8 are clearly sequenced with goals, explicit STOP points after each step, a MANDATORY 300-row gate with a poll-until-READY feedback loop, error-recovery guidance for build and runtime failures, and a user-confirmation gate before destructive DROP statements. A 'What Remains Active' table and consolidated stopping-points checklist close the loop.	5 / 5
Progressive Disclosure	Section headers and a project-structure map give reasonable navigation, and no bundle files exist so nothing is mis-referenced. But large reference material is inlined in a single 700-line file — the semantic-view DDL, the 'Key Consumer Patterns (Reference)' Java/API material, and three near-duplicate forecast recipes all belong in one-level-deep reference files per the rubric's rationale that SKILL.md should be an overview.	3 / 5
	Total	16 / 20 Passed

Description

87%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is a strong example: concrete third-person actions, a specific technology niche, and both an explicit 'Use when' clause and a trigger keyword list. The only weakness is that it undersells the skill's full scope — the ML forecasting and semantic-view analytics that make up roughly half the body get no mention.

Dimension	Reasoning	Score
Specificity	"Set up, configure, run, and debug" names four concrete actions against specific technology (custom Kafka consumer, Snowflake, Snowpipe Streaming SDK v2). Not a 5 because the actions cover only the consumer pipeline and give no hint of the skill's analytics half (ML FORECAST, semantic views), a minor coverage gap.	4 / 5
Completeness	Explicitly answers both questions: what — "Set up, configure, run, and debug a custom Kafka consumer that streams data into Snowflake via Snowpipe Streaming SDK v2"; when — "Use when: building a Kafka-to-Snowflake streaming pipeline, running the CDR demo, or troubleshooting the custom consumer" plus a dedicated "Triggers:" list of concrete phrases. Third-person voice throughout.	5 / 5
Trigger Term Quality	"kafka consumer, kafka snowflake, snowpipe streaming kafka, CDR demo, kafka to snowflake, custom consumer, streaming ingest kafka" gives good natural keyword coverage. Not a 5 because a few natural variations users would say are missing (e.g. "call detail records", "snowflake kafka connector", "cdr").	4 / 5
Distinctiveness Conflict Risk	"Snowpipe Streaming SDK v2" and "CDR demo" define a clear niche with distinct triggers ("kafka snowflake", "snowpipe streaming kafka", "CDR demo") that no generic Kafka or Snowflake skill would claim, so conflict risk is minimal.	5 / 5
	Total	18 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (706 lines); consider splitting into references/ and linking	Warning

	Total	15 / 16 Passed

Repository: snowflakedb/snowpipe-streaming-sdk-examples
Path: custom-kafka-consumer/.cortex/skills/custom-kafka-consumer/SKILL.md
Commit: 0c5fd89

Reviewed: 7 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.