skilld
All skills
kotlin avatar

/kotlin-tooling-kotlin-toolchain

@c2f9069 official
by kotlinkotlin/kotlin-agent-skills1.1k stars
42

Load when building, running, testing, packaging, linting, or configuring a Kotlin/Java project with the Kotlin Toolchain (JetBrains' unified CLI, formerly Amper), when scaffolding a new or greenfield Kotlin project, or when the repo has project.yaml, module.yaml, or a ./kotlin wrapper. Skip for existing Gradle/Maven projects.

Ask your Agent

Use this Skill: https://skilld.dev/gh/kotlin/kotlin-agent-skills/kotlin-tooling-kotlin-toolchain

This session only. Nothing lands on disk.

evalsEVALUATION.md

≈466 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Evaluation

This skill is A/B evaluated with a scaffolding task pair: arm A scaffolds a Kotlin project with Android and iOS platforms without the skill, and arm B runs the identical prompt with the kotlin-tooling-kotlin-toolchain skill mounted. Both arms run the Kotlin Toolchain 0.11.1 CLI, and the verifier scores whether the result is a working Toolchain project (project.yaml + module.yaml, ./kotlin wrapper, no Gradle build) that loads and builds.

The machine-checkable expectations are captured in evals.json.

Latest results (2026-08-14)

Run with the JetBrains skills A/B eval harness, claude-code agent, n = 10 pairs per model.

Model Reward without skill Reward with skill Δ Significance
anthropic/claude-opus-5 0.00 ± 0.00 0.83 ± 0.00 +0.83 p < 0.05
anthropic/claude-sonnet-5 0.00 ± 0.00 0.83 ± 0.00 +0.83 p < 0.05

Without the skill both models default to Gradle (6–7 gradle invocations per trial, kotlin init never used) and score zero on the Toolchain rubric. With the skill they switch to the Toolchain (0 gradle calls, 6–7 kotlin calls, kotlin init used in every trial) and the reward is near-deterministic (σ ≈ 0.00): the build-tool decision moves into the skill rather than the model's reasoning. The with-skill arm also uses markedly fewer tool calls and tokens.

Scope note

The A/B suite above targets the kotlin-tooling-kotlin-toolchain skill directly. The companion skills (kotlin-tooling-kotlin-toolchain-plugin-authoring, kotlin-tooling-gradle-to-kotlin-toolchain-plugin, kotlin-tooling-gradle-to-kotlin-toolchain-project) share the same Toolchain 0.11.1 baseline and model/agent, and ship their own evals.json; their formal A/B runs are tracked separately.

Source: SKILL.md on GitHub

Third-party checks

No alerts20d3 checks · Risk SAFE
  • Gen Agent Trust Hub20d

    The skill provides instructions for using the JetBrains Kotlin Toolchain for project scaffolding, building, and testing. It contains legitimate commands and references to official JetBrains resources and SDKMAN, while also including explicit security recommendations for the agent to handle untrusted project inputs and wrapper scripts safely.

  • Socket20d

    No alerts

  • Snyk20d

    Risk: LOW · No issues

Provenance

Signed by skilld at c2f9069. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 3 weeks ago.

Activeupdated last month

Capability

Other metadata
metadata
{
  "author": "github:@singleton11",
  "version": "0.1.0",
  "provider": "anthropic",
  "tested_models": "anthropic/claude-sonnet-5, anthropic/claude-opus-5",
  "agent_version": "claude-code",
  "last_eval": "2026-08-14"
}

README badge

README badge for kotlin/kotlin-agent-skills/kotlin-tooling-kotlin-toolchain