---
name: debate_skills
description: Prepares competitive debate material and coaching. Builds cases with contentions and impacts, drafts rebuttals, builds flow sheets, and explains how rounds are judged across LD, Public Forum, Policy, and APDA parliamentary formats. Triggers when the user asks to prep a case, write or rebut an argument, flow a round, or asks about debate rules, formats, or judging criteria.
---

# Debate Skills

## Purpose

Helps a user (or an agent) prepare for, argue in, and evaluate competitive debate rounds. Covers the four jobs in the brief: building a case, rebuttal, flowing arguments, and how rounds are judged. Formats covered: Lincoln-Douglas (LD), Public Forum (PF), Policy, and APDA parliamentary. Speech orders and times live in `references/format-times.md`.

## Workflow

1. **Identify the format and side.** If the user did not say, ask which format (LD, PF, Policy, APDA) and which side (affirmative/pro/gov vs. negative/con/opp). Everything downstream depends on this.
2. **Build the case.** Every contention follows claim, warrant, evidence, impact. Two or three contentions per case. LD adds a value and a value criterion (the yardstick the judge uses to measure the value). PF adds uniqueness: establish the current state of the world before arguing it must change.
3. **Draft the rebuttal.** For each opposing argument: signpost its location on the flow, attack the link, the warrant, or the evidence, then weigh your impact against theirs. Answer their strongest arguments first.
4. **Flow the round.** Track each speech's arguments in columns and flag drops. See `references/flowing.md`.
5. **Collapse and crystallize.** Summaries and final focuses (or final rebuttals) narrow to the 2-3 voting issues you are winning and weigh them comparatively. Never introduce new arguments there.
6. **Judge.** Score by the format's criteria and write the ballot as an RFD: decision, voting issues, analysis, weighing. See `references/judging-criteria.md`.

## Output Contract

- **Case:** framework (LD: value + criterion), 2-3 contentions each with a tag line, warrant, cited evidence or real examples, and a stated impact. Evidence carries dates; never invent them. Where the user has no cards, deliver a skeleton and mark where evidence is needed.
- **Rebuttal:** per-argument responses with signposts, plus guidance on which responses to extend into later speeches.
- **Flow sheet:** a grid mapping each speech's arguments to responses, with dropped arguments flagged.
- **Rules/judging answers:** name the format, the criteria, and the speaking order. Cite sources from the notes when possible; never invent a rule.

## Operating Rules

- Every factual claim in output must be verified or framed as opinion. Never invent numbers, dates, quotes, studies, tournaments, or results.
- Silence equals concession. An argument you do not answer in your next speech is treated as conceded, so rebuttals answer the opponent's strongest points first.
- Signpost everything. Before addressing an argument, say where you are: "On their Contention 2...", "Back on my case..."
- Weigh comparatively, using magnitude, probability, timeframe, scope, and reversibility. Restating both impacts side by side is not weighing.
- PF rule (NSDA unified manual): no plans or counterplans. If the user's idea is a formalized proposal, reframe it as advocacy reasoning.
- LD rule: no plan arguments from either debater in most league play; check the local rules.
- Judge the arguments made in the round, not personal beliefs about the topic. If an argument is unclear or unexplained, you cannot vote on it.
- Worked examples live in `references/worked-examples.md`. Follow them when drafting; they show the structure, not canned content.