Koog 1.0 idioms, gotchas, and scaffolding skills for Kotlin agents on the JVM
88
88%
Does it follow best practices?
Impact
88%
1.95xAverage score across 43 eval scenarios
Passed
No known issues
{
"context": "Negative case — tests whether the agent recognizes the developer's workflow has a known compile-time topology (three fixed phases, LLM only picks tools within each phase) and routes them to the graph DSL instead of building a planner. A planner here would be over-engineering: extra LLM round-trips for planning decisions the developer already made, with no autonomy benefit.",
"type": "weighted_checklist",
"checklist": [
{
"name": "Names the topology as the disqualifying signal",
"description": "Identifies in prose that the developer's workflow has a fixed phase ordering known at design time — not that the LLM picks tools within phases. Surfaces the test from the Koog tile: the planner is for runtime-decided sequencing, not for tool selection within fixed steps",
"max_score": 25
},
{
"name": "Redirects to the graph DSL",
"description": "Recommends authoring a graph strategy (or invokes the appropriate sibling skill) instead of installing the planner module. Names the strategy DSL specifically — not a vague \"different approach\"",
"max_score": 30
},
{
"name": "Does not install agents-planner",
"description": "Does NOT add ai.koog:agents-planner to build.gradle.kts and does NOT produce Planners.llmBased / Planners.goap factory calls. Installing the wrong module on the developer's say-so without surfacing the mismatch is the failure mode this scenario tests",
"max_score": 20
},
{
"name": "Acknowledges the developer's framing without capitulating",
"description": "Engages with the developer's stated reason (\"letting the agent plan its work sounded right\") and explains the distinction between LLM-picks-tools (graph) and LLM-picks-step-sequence (planner). Does not just refuse without explanation",
"max_score": 15
},
{
"name": "Mentions the cost penalty of an unwarranted planner",
"description": "Notes that a planner adds an LLM round-trip per planning decision — for a fixed three-phase workflow that's wasted cost with no autonomy benefit. The developer needs this to internalize the recommendation",
"max_score": 10
}
]
}evals
scenario-1
scenario-2
scenario-3
scenario-4
scenario-5
scenario-6
scenario-7
scenario-8
scenario-9
scenario-10
scenario-11
scenario-12
scenario-13
scenario-14
scenario-15
scenario-16
scenario-17
scenario-18
scenario-19
scenario-20
scenario-21
scenario-22
scenario-23
scenario-24
scenario-25
scenario-26
scenario-27
scenario-28
scenario-29
scenario-30
scenario-31
scenario-32
scenario-33
scenario-34
scenario-35
scenario-36
scenario-37
scenario-38
scenario-39
scenario-40
scenario-41
scenario-42
scenario-43
skills
add-observability
add-persistence
add-rag
add-structured-output
add-token-budgeting
add-tool
cache-llm-calls
define-prompt
domain-model-subtask-pipeline
references
enable-prompt-caching
handle-agent-events
manage-state
migrate-from-0-x
model-planner-subtasks
persist-chat-history
query-sql-from-agent
scaffold-agent
snapshot-and-restore
test-koog-agents
trace-agent-internals
use-attachments
use-functional-agent
use-llm-node-variants
use-planner
wire-a2a
wire-acp-server
wire-ktor-server
wire-mcp-server
wire-spring-boot