auto-kaggle

Name: auto-kaggle
Author: zc6600

Procedure for autonomous Kaggle iteration with guarded submissions.

by zc6600(0) 0 installs

Free

Free account. Installing gives you the manifest plus copy-paste snippets.

See reviews

About

Imported from zc6600/aura (use-cases/auto-kaggle/template/skills/auto-kaggle/SKILL.md). Install upstream with npx skills add zc6600/aura --skill auto-kaggle. Copyright stays with the author.

AutoKaggle Skill

Operating Rules

Always read params/autokaggle.yml first.
Treat the experiment registry as the source of truth.
Do not submit directly with shell or raw Kaggle CLI.
Before real submit, make sure verifier evidence is recorded.
If submit guard returns wait_required, call timer for wait_chunk_seconds.
After waiting, retry the same candidate.
After real submit, poll Kaggle submissions and record feedback.

Candidate Requirements

Each candidate must have:

run_id
hypothesis
local CV score
submission path
registry record
verifier result before real submit
submit guard result
leaderboard feedback or dry-run feedback

Use it

Copy one of these into your project. Installing also returns the manifest and these snippets.

# after Install: the listing is in your workspace's routing pool - a plan picks it for its slot
curl -s -X POST https://api.opensmartroute.ai/api/v1/route -H 'Authorization: Bearer $OSR_API_KEY' -H 'Content-Type: application/json' -d '{"text": "...", "plan": true}'

Installed into a catalogue, chosen by a router

Install auto-kaggle and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

Photo: RDNE Stock project

Manifest

An Open Capability Manifest: the router reads it to know what this does, what it costs and when to pick it.

zc6600-aura-auto-kaggle.ocm.jsonjson

{
  "ocm": "1",
  "id": "zc6600-aura-auto-kaggle",
  "kind": "skill",
  "name": "auto-kaggle",
  "description": "Procedure for autonomous Kaggle iteration with guarded submissions.",
  "publisher": "zc6600",
  "version": "1.0.0",
  "capabilities": {
    "domains": [
      "general"
    ],
    "tags": [
      "skill-md",
      "github"
    ],
    "languages": [
      "en"
    ]
  },
  "quality_prior": 0.6,
  "examples": [
    "Procedure for autonomous Kaggle iteration with guarded submissions."
  ],
  "primary": false,
  "metadata": {
    "source": {
      "provider": "github",
      "repository": "https://github.com/zc6600/aura",
      "path": "use-cases/auto-kaggle/template/skills/auto-kaggle/SKILL.md",
      "ref": "4927de1d03e2be2763c197a6e9d0ff0368578bc7",
      "url": "https://github.com/zc6600/aura/blob/4927de1d03e2be2763c197a6e9d0ff0368578bc7/use-cases/auto-kaggle/template/skills/auto-kaggle/SKILL.md",
      "key": "zc6600/aura/use-cases/auto-kaggle/template/skills/auto-kaggle/SKILL.md"
    }
  },
  "instructions": "# AutoKaggle Skill\n\n## Operating Rules\n\n- Always read `params/autokaggle.yml` first.\n- Treat the experiment registry as the source of truth.\n- Do not submit directly with shell or raw Kaggle CLI.\n- Before real submit, make sure verifier evidence is recorded.\n- If submit guard returns `wait_required`, call `timer` for `wait_chunk_seconds`.\n- After waiting, retry the same candidate.\n- After real submit, poll Kaggle submissions and record feedback.\n\n## Candidate Requirements\n\nEach candidate must have:\n\n- `run_id`\n- hypothesis\n- local CV score\n- submission path\n- registry record\n- verifier result ",
  "cost": {
    "context_tokens": 170
  }
}

Fetch it by URL: GET /api/v1/registry/zc6600-aura-auto-kaggle/manifest?version=1.0.0