sakanaai/ai-scientist-v3/create-idea

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Create one next-step idea bundle and matching rubric bundle for iterative optimization. Use during idea generation runs (for example via scripts/idea.sh) when `idea_bundle.json` and `target_rubrics.json` must be produced.

Install

npx vskill@latest install sakanaai/ai-scientist-v3/create-idea

SKILL.md

What the skill is about

Produce one concrete, testable next experiment idea with machine-readable rubrics.

Output creation

Write both required JSON files directly with the available file-writing tools. Do not depend on command execution. When command execution is explicitly available, this optional helper can create starter files:

bash
# the script path is relative to the skill file
python ./scripts/create_idea_placeholder.py <working_dir> [--cycle-index N]
  • <working_dir>: the agent's current working directory (idea output directory).
  • --cycle-index N: cycle index for the idea_id prefix (falls back to CYCLE_INDEX env var, then defaults to 1).

The optional script creates idea_bundle.json and target_rubrics.json with TODO placeholders for every required field. Replace each TODO value with the actual content.

What to do

  1. Read task context:
  • {{TASK_MD_PATH}}
  • {{TASK_GOAL}}
  • {{TASK_RUBRICS}}
  1. Use cycle and history context:
  • cycle_index={{CYCLE_INDEX}}
  • recent ideas and attempts from the prompt context
  • best-known manifest from the prompt context
  1. Write exactly these files directly (or use the optional placeholder helper when execution is allowed):
  • {{IDEA_BUNDLE_FILE}}
  • {{TARGET_RUBRICS_FILE}}
  1. {{IDEA_BUNDLE_FILE}} must be a JSON object with keys:
  • idea_id (string)
  • title (string)
  • hypothesis (string)
  • implementation_plan (non-empty array of short strings)
  • target_rubrics (non-empty array of {id, description, type})
  • notes_for_agent (string)
  1. Rubric rules:
  • include at least one metric rubric
  • include at least one implementation rubric
  • use stable snake_case rubric ids
  • keep rubric descriptions concrete and measurable
  1. Ensure {{TARGET_RUBRICS_FILE}} is exactly equal to idea_bundle.json.target_rubrics.
  2. Apply the active harness rules exactly:

{{HARNESS_RUBRIC_REQUIREMENTS}}

When a rubric includes verification, verification must be a JSON object, never a prose string. For harness implementation rubrics, use exactly this shape:

json
{
"verification": {
"type": "source_contains",
"patterns": ["specific executable snippet"]
}
}
  1. Before finishing, reread both files and confirm every implementation rubric satisfies the active harness rules.
  2. Do not modify files outside the required output files.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/sakanaai/ai-scientist-v3/create-idea/badge)](https://verified-skill.com/skills/sakanaai/ai-scientist-v3/create-idea)

Something wrong with this skill or its page? Tell us.