Keep guideline notes searchable, draft worker-facing task packages, and check onboarding-style questions without leaving the browser.
Library — Save any guideline docs, rubrics, or rejection rules here. Give each one a title and tags, then paste the text. Everything you save becomes searchable from the bar above — the tool pulls matching rules when you analyze a question or build a prompt package.
Guideline Library
0 matches
Prompt Maker — Pick a domain, expertise level, and task shape, then hit Generate Draft to auto-fill all fields with a scenario-based starter. Replace any generic details with your real source material, then hit Build Package to produce the formatted submission fields ready to paste one-by-one into the Outlier TBench form.
Generate a worker submission draft, replace any generic details with real source material, then use the checks to catch weak resources, vague outputs, or unverifiable grading.
Generated Submission Package
Fill the fields and build a task package.
LLM / Reviewer Risk ChecksReadiness DashboardCross-Field ConsistencyGuideline ChecksSubmission Field Audit
Pipeline Runner
Disconnected
Checking backend at http://127.0.0.1:8787…
Enterprise Pipeline
Use step 3 in the top workflow row.
Status
Execution Status:NOT RUN
Task Family:—
solve.py:—
verify.py:—
outputs/:—
placeholder scan:—
copy gate:LOCKED
Actions
Execution Logs
No logs yet.
Answer Helper — Paste an onboarding question or scenario into the box and hit Analyze. The tool searches your saved guidelines for matching rules and drafts a structured answer outline. Requires at least one saved guide in the Library to work.
Answer Helper
Relevant Guidelines
Add guidelines, then analyze a question.
Answer Outline
The outline will appear here.
Code Templates — Starter scaffolds for the domain and scenario you last generated in Prompt Maker. These are not finished solutions — fill every TODO, replace all placeholder paths with your real data paths, and drop them into your zip as the basis for the solve.py and verify.py you actually submit.
Code Templates
generate a draft first
Starter scaffolds for the domain and scenario last generated in Prompt Maker. Fill every TODO, replace placeholder paths with your real data paths, and drop these into your zip as the basis for the actual solve.py and verify.py you submit.
solve.py
Go to Prompt Maker, hit Generate Draft, then come back here.
verify.py
Go to Prompt Maker, hit Generate Draft, then come back here.