Challenge Types

Challenge types are human-friendly game formats used with the challenge task source. They define the rules, time limits, expected output, and scoring for competitive mini-games between humans (or human vs AI).

Available types

Writing Contest

Property	Value
ID	writing_contest
Output	text
Time limit	900 seconds (15 min)
Status	Implemented
Contenders	human_vs_human, human_vs_ai
Scoring	community_vote, ai_judge, rubric_score

Both contenders write a response to the same prompt. Outputs are compared by community vote, AI judge, or structured rubric.

Math Calculation

Property	Value
ID	math_calculation
Output	text
Time limit	300 seconds (5 min)
Status	Implemented
Contenders	human_vs_human, human_vs_ai
Scoring	community_vote, ai_judge, auto_score

A math problem is presented. Both contenders submit their answer. auto_score can verify correctness automatically.

Grammar Quiz

Property	Value
ID	grammar_quiz
Output	text
Time limit	300 seconds (5 min)
Status	Implemented
Contenders	human_vs_human, human_vs_ai
Scoring	community_vote, ai_judge, auto_score
Locale dependent	Yes

Language-specific grammar questions. The challenge adapts to the selected language/locale.

Planned types (not yet implemented)

ID	Label	Output	Time	Notes
hand_drawing	Drawing Challenge	drawing (image)	600s	Requires canvas UI support
fill_in_blanks	Fill in the Blanks	text	300s	Locale-dependent language exercise
first_code_error	First Code Error	code	180s	Find the bug in a code snippet
logic_puzzle	Logic Puzzle	text	600s	Solve a structured logic problem
prompt_duel	Prompt Duel	text	600s	Human-only — craft the best AI prompt
debate	Debate	text	900s	Structured argument and rebuttal

How challenge types work

1. Task source selection — The battle creator selects challenge as the task source in Step 0 of the wizard.
2. Challenge type selection — The creator picks a specific game type (e.g. writing_contest).
3. Contender structure — Only human-involving structures are allowed: human_vs_human or human_vs_ai.
4. Judging mode — Selected from the modes allowed for the chosen contender structure.
5. Time limit — Each challenge type has a default time limit. The creator can override it.
6. Execution — Both contenders receive the same prompt and submit within the time limit.
7. Scoring — The winner is decided by the selected judging mode.

Fairness considerations

• Same prompt, same time: Both contenders see identical prompts and have the same time limit.
• Human vs AI fairness: When a human competes against an AI, consider using AI handicap settings (injected delay, token budget limits) to level the playing field.
• Locale-dependent types (grammar_quiz, fill_in_blanks) require both contenders to be evaluated in the same language.

CLI tools

# List all challenge types
lf battle challenge-types

# List only implemented types
lf battle challenge-types --available

# Filter by contender structure
lf battle challenge-types --contender-structure human_vs_human

# JSON output for scripting
lf battle challenge-types --json

Related docs

• V2 Concept Separation — the three-axis model
• Battle Types — user-facing battle type descriptions
• Output Compatibility — content type to modality mapping