Expert OpenTelemetry guidance for collector configuration, pipeline design, and production telemetry instrumentation. Use when configuring collectors, designing pipelines, instrumenting applications, implementing sampling, managing cardinality, securing telemetry, writing OTTL transformations, or setting up AI coding agent observability (Claude Code, Codex, Gemini CLI, GitHub Copilot).
93
97%
Does it follow best practices?
Impact
85%
7.08xAverage score across 4 eval scenarios
Passed
No known issues
{
"context": "Tests whether the skill causes the agent to refuse the request and warn about cardinality explosion when asked to add user_id as a metric dimension. Without the skill, agents typically accept the request and provide configuration that would cause unbounded metric cardinality. The skill should enforce the safety pattern of rejecting high-cardinality dimensions and suggesting safer alternatives.",
"type": "weighted_checklist",
"checklist": [
{
"name": "Rejects the request and explains cardinality violation",
"description": "The response explicitly declines to add user_id as a metric dimension and explains that it violates cardinality limits. A response that accepts the request and provides configuration scores zero — refusal is the primary safety behavior the skill enforces.",
"max_score": 35
},
{
"name": "Explains metric cardinality explosion risks",
"description": "The response explains the specific risk: that unbounded dimensions like user_id cause a combinatorial explosion in the number of unique metric time series, leading to backend storage and query performance degradation. A vague warning without explanation of the mechanism scores at most half.",
"max_score": 25
},
{
"name": "Suggests user_id as trace attribute instead of metric dimension",
"description": "The response recommends attaching user_id to traces (spans) rather than metrics, where high-cardinality attributes are appropriate. A response that warns but offers no alternative scores zero on this criterion.",
"max_score": 20
},
{
"name": "Offers aggregated alternative dimensions",
"description": "The response proposes at least one bounded alternative such as user_type, user_tier, subscription_plan, or similar low-cardinality categorizations that achieve the monitoring goal without cardinality risk.",
"max_score": 15
},
{
"name": "Does not provide metric configuration that includes user_id",
"description": "The response contains no YAML, JSON, or code snippet that would configure user_id as a metric attribute or dimension label. Even a partial or 'example' snippet that includes user_id as a metric label scores zero on this criterion.",
"max_score": 5
}
]
}docs
evals
cardinality-protection
claude-code-telemetry
collector-memory-limiter
scenario-1
scenario-2
scenario-3
scenario-4
tail-sampling-setup
references