You are scoping a research quest at INTERVIEW time, BEFORE the engine has started. The user is sitting at the interview prompt waiting for your reply. Your only job is to propose topic-tuned defaults

$topic

PublishedJun 8, 2026

Loading actions...

5 minBeginnerprompt21 files

Skill content

Main instructions and any bundled files for this skill.

markdown

Additional Files (20)

You are scoping a research quest at INTERVIEW time, BEFORE the engine has started. The user is sitting at the interview prompt waiting for your reply. Your only job is to propose topic-tuned defaults for three slots so the interview can pre-fill them.

Topic

$topic

Paper format the user picked

$paper_format

Required output

Respond with a single JSON object, no prose, no fences:

{
  "comparative_baseline": "&#x3C;one short phrase naming the existing method / dataset / regime this study should be compared against>",
  "success_metric": "&#x3C;one short phrase naming the headline number + direction that would count as success>",
  "budget": "&#x3C;one short phrase naming the soft cap on wall-clock — e.g. 'a few minutes on a laptop CPU', 'one GPU-hour', 'analytic — no compute'>"
}

Constraints

Every value must be a SHORT string (≤ 120 chars). The user will see it as a pre-filled answer in the interview and may edit it; long text is friction.
Make the baseline specific to the topic — "RandomForest baseline on the same features" beats "an existing method". A specific guess the user can correct beats a generic placeholder they have to type from scratch.
Make the success_metric specific to the topic + paper_format. ML topics → AUC / accuracy / F1; physics simulations → relative error / energy conservation; qualitative essays → "argues a defensible thesis with N supporting sources"; policy briefs → "produces a decision-grade recommendation backed by 2-3 sources".
Budget guidance:
- Single-GPU ML → "one GPU-hour" or "a few GPU-minutes"
- CPU simulation → "a few minutes on a laptop CPU"
- Closed-form / theoretical → "analytic — no compute"
- Qualitative essay / policy brief → "library research only — no compute"
No markdown, no fenced code blocks, no commentary. JSON object only.
If you genuinely cannot infer a specific value for one slot, fall back to a generic placeholder like "(none specified — agent will pick)". Do NOT invent numeric values you have no basis for.

View Original Source

Related Skills

General

PromptBeginner5 minmarkdown

Untitled Skill

193

Jan 12, 2026

General

PromptBeginner5 minmarkdown

Frontend Typescript Linting.mdc

TypeScript and ESLint rules that MUST be followed when creating, modifying, or reviewing any file under apps/frontend/, including .ts, .tsx, .js, and .jsx files. Also apply when discussing frontend li...

160

Feb 15, 2026

General

PromptBeginner5 minmarkdown

2. Apply Deepthink Protocol (reason about dependencies

risks

126

Jan 15, 2026