agentleFS
Sign inStart free
cukascukas/claudes-ai-buddies/skills/tribunal/SKILL.md

tribunal

Skill9 starsChanged 8 months agoMITView on GitHub

Multi-mode AI debate — adversarial, Socratic, steelman, red-team, synthesis, or postmortem with evidence

12 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use production-grade, session-handoff and 1 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Sean-MichaelSkill

    production-grade

    Apply production-grade standards when building or modifying web applications. Trigger on: "make it production ready", "add auth", "production grade", "harden this", or any time a user is building a web app with user accounts, an API, or a database and hasn't said it's a prototype. Standards are the default — not an add-on.

    Also in 3 projects that use this

  • wpankSkill

    session-handoff

    WHAT: Create comprehensive handoff documents that enable fresh AI agents to seamlessly continue work with zero ambiguity. Solves long-running agent context exhaustion problem.

    Also in 3 projects that use this

  • SentrySkill

    skill-creator

    Create new agent skills following the Agent Skills specification. Use when asked to "create a skill", "add a new skill", "write a skill", "make a skill", "build a skill", or scaffold a new skill with SKILL.md. Guides through requirements, writing, registration, and verification.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 10 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Parse the question and mode: Check for mode flags (--socratic, --steelman, --red-team, --synthesis, --postmortem, or --mode X).

    line 37
  2. Very specific:

    Before dispatching: Use your judgment — if the conversation contains context that would help the debaters (constraints, failed approaches, …

    line 50
  3. Distinctive:

    Read the file matching the mode from the manifest, then follow its judging/synthesis instructions and output format.

    line 86
  4. Distinctive:

    Forge close call: When forge scores are within 3 points

    line 101
  5. Very specific:

    Review disagreement: When /codex-review and /gemini-review conflict

    line 102
  6. Distinctive:

    Evidence over eloquence: Every claim needs file:line evidence.

    line 113
  7. Distinctive:

    Verify citations: Always read referenced files to confirm.

    line 114

“Common” is measured: kept by at least 6 in 10 of the 10 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/tribunal — Multi-Mode AI Debate

Two AI buddies engage on a codebase question using one of six modes. Every claim requires FILE:LINE evidence.

Modes

ModeFlagAIs doClaude's roleBest for
adversarial(default)Argue FOR vs AGAINSTJudge — picks winnerBinary decisions, should/shouldn't
socratic--socraticProbe assumptions with questionsSynthesizer — surfaces insightsEarly exploration, unclear framing
steelman--steelmanArgue the OTHER side's strongest caseCalibrator — shows true strength of each sideAvoiding confirmation bias
red-team--red-teamAttack from different angles, no defenseRisk assessor — prioritized vulnerability tablePoking holes in designs/plans
synthesis--synthesisEach proposes a solution, then hybridizeMerger — evaluates proposals + hybridFinding a third option
postmortem--postmortemInvestigate failure from different anglesInvestigator — unified timeline + root causeBug investigation, incident analysis

How to invoke

/tribunal "Should we refactor the auth middleware?"
/tribunal --socratic "Is our error handling resilient enough?"
/tribunal --steelman "Should we migrate to microservices?"
/tribunal --red-team "Review our new payment flow"
/tribunal --synthesis "How should we restructure the data layer?"
/tribunal --postmortem "Why did the deploy fail yesterday?"
/tribunal --mode red-team "Audit the new API endpoints"

Step-by-step workflow

Phase 0: Setup

  1. Parse the question and mode. Check for mode flags (--socratic, --steelman, --red-team, --synthesis, --postmortem, or --mode X). Default: adversarial.
  2. Detect available buddies:
source "${CLAUDE_PLUGIN_ROOT}/hooks/lib.sh"
AVAILABLE=$(ai_buddies_available_buddies)
  1. Require at least 2 buddies.
  2. Tell the user which buddies will participate, the mode, and round count.

Phase 1: Dispatch

Before dispatching: Use your judgment — if the conversation contains context that would help the debaters (constraints, failed approaches, user preferences), summarize it in $CONVERSATION_CONTEXT. If the question is self-contained, leave it empty.

IMPORTANT: Set the Bash tool's timeout parameter to 600000 (10 minutes). The default Bash timeout (120s) will kill slow engines mid-debate.

MANIFEST_PATH=$(bash "${CLAUDE_PLUGIN_ROOT}/scripts/tribunal-run.sh" \
  --question "THE_QUESTION" \
  --cwd "$(pwd)" \
  --mode MODE_NAME \
  --rounds 2 \
  --timeout 600 \
  --conversation-context "$CONVERSATION_CONTEXT")

Phase 2: Read results

Read $MANIFEST_PATH (tribunal-manifest.json):

{
  "question": "...",
  "mode": "adversarial|socratic|steelman|red-team|synthesis|postmortem",
  "rounds": 2,
  "debaters": ["codex", "gemini"],
  "arguments": { ... }
}

Phase 3: Evaluate

Read the mode-specific guide for how to evaluate results and format output:

${CLAUDE_PLUGIN_ROOT}/skills/tribunal/modes/{mode}.md

Read the file matching the mode from the manifest, then follow its judging/synthesis instructions and output format.

Round constraints

ModeRoundsReason
adversarialFlexible (default 2)More rounds = deeper rebuttals
steelmanFlexible (default 2)Like adversarial but with reversed positions
socraticFixed 2Round 1 asks, Round 2 answers
red-teamFixed 2Round 1 attacks, Round 2 chains attacks
synthesisFixed 2Round 1 proposes, Round 2 hybridizes
postmortemFixed 2Round 1 investigates, Round 2 cross-examines

Auto-triggers

  1. Forge close call: When forge scores are within 3 points
  2. Review disagreement: When /codex-review and /gemini-review conflict

Configuration

KeyDefaultDescription
tribunal_rounds2Default round count
tribunal_max_buddies3Max debaters

Rules

  • Evidence over eloquence. Every claim needs file:line evidence.
  • Verify citations. Always read referenced files to confirm.
  • No-evidence = zero. Enforce in all modes.
  • Follow the mode guide. Each mode has different judging criteria and output format.
  • Always clean up worktrees after the session.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free