verdict

Verdict

Context: $ARGUMENTS

Deliver the final judgment. What survives, what doesn't, and the uncomfortable truth.

Workflow

1. Check for existing graph state

uv run ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py show

If a graph exists from prior /steelman or /crucible runs, use it. If empty, build one from scratch by identifying the idea's claims, assumptions, evidence, weaknesses, and counter-arguments:

uv run ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py reset
uv run ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py add "<claim-id>" claim --obs "<fact>"
uv run ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py add "<assumption-id>" assumption --obs "<fact>"
uv run ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py add "<weakness-id>" weakness --obs "<fact>"
uv run ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py relate "<claim-id>" assumes "<assumption-id>"

2. Gadfly sting

Read ${CLAUDE_SKILL_DIR}/references/social-gadfly.md.

Name the ONE thing nobody wants to say about this idea. Not a weakness (those are tactical), the foundational uncomfortable truth that would remain even if everything else went right.

Ask yourself:

• What is everyone avoiding saying?
• What assumption is so foundational that questioning it feels rude?
• What would a persistent, unsilenceable critic keep coming back to?

If the gadfly sting reveals something that changes the structural picture, register it:

uv run ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py add "<counter-id>" counter --obs "<the uncomfortable truth>"
uv run ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py relate "<counter-id>" undermines "<claim-id>"

3. Structural verdict

Run the graph verdict:

uv run ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py verdict

This returns: claims that survive scrutiny (supported, no unmitigated undermining), claims that don't survive (depend on contradicted or ungrounded assumptions), claims needing more work (contested, mixed evidence), and unmitigated weaknesses.

4. Present the verdict

Output format

The Uncomfortable Truth: One sentence. The gadfly sting.

Survives Scrutiny: Claims with support and no unmitigated undermining. These are the parts worth keeping.

Does Not Survive: Claims depending on contradicted or ungrounded assumptions. These need to be dropped or fundamentally reworked.

Needs More Work: Contested claims where evidence is mixed. Not dead, not proven. What additional evidence or testing would resolve them.

Unmitigated Weaknesses: Exposed risks with no defense. What needs to happen before this idea can move forward.

Optionally export the full graph for the user:

uv run ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py export

Resources

• ${CLAUDE_SKILL_DIR}/references/social-gadfly.md — load always. Socrates as the stinging fly, the cost of silencing uncomfortable questioners, the distinction between tactical weaknesses and foundational truths.
• ${CLAUDE_SKILL_DIR}/references/polemic.md — load when the gadfly sting needs rhetorical force or the user asks for a forceful take. Constructive polemic exposes genuine errors without demonizing.
• ${CLAUDE_PLUGIN_ROOT}/scripts/argument-graph.py — run verdict for structural analysis, export for full graph markdown.

Gotchas

• The gadfly sting is not another weakness. Weaknesses are tactical (the API might go down). The gadfly names a foundational truth (the team doesn't actually want this to succeed). If your gadfly sting could appear in a list of weaknesses, it's not stinging enough.
• If a graph exists from prior skills, use it. The verdict is strongest when it reads a graph built by /steelman and /crucible. Running verdict standalone works, but the graph will be thinner.
• The verdict script's output is structural, not rhetorical. It tells you what survives based on graph topology. Your job is to synthesize that into a human-readable judgment.
• Don't soften the gadfly. Socrates was executed for it. The uncomfortable truth is uncomfortable. Name it plainly.