roundtable

A structured discussion between several specialized agents about a development decision.

In plain words
What is it for?
Use it for choices involving competing concerns or multiple areas of expertise, such as product scope, design, and engineering trade-offs.
Why use it?
It gathers different viewpoints, makes disagreements explicit, and lets a human approve the resulting decisions.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/trustmybot/plugin/roundtable
Clone the repo
git clone --depth 1 https://github.com/trustmybot/plugin
Per session 15 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 636 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00015 $0.00636
Opus 5 $0.00008 $0.00318
Sonnet 5 $0.00003 $0.00127
Haiku 4.5 $0.00002 $0.00064

Measured yesterday against content hash b41ec380456d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

roundtable scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

commands/roundtable.md · 55 lines

How it starts

The opening of the file, as written. The whole thing — 55 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Roundtable on: $ARGUMENTS

Run a structured deliberation: collect parallel agent positions, synthesize into agreements and disagreements, ratify with the Human, then close with decisions.

If $ARGUMENTS is empty, ask the user for the topic via AskUserQuestion before proceeding.

When to invoke

  • Divergent opinions, multi-dimension trade-offs, cross-discipline calls.
  • Skip for: factual lookups, single-discipline decisions, or delegation.

Phase 1 — Setup

Glob .claude/agents/ and pick the 2–4 agents whose domain bears on the topic, excluding SWE; halt if fewer than 2 are relevant.

Resolve the carrier issue — the issue this roundtable's discussions and decisions attach to. If the topic already maps to an open issue, use its issue_id. Otherwise create a one-shot carrier with issue_create (title = the topic) and use the returned id; Phase 6 closes it once the roundtable ends.

Open the round with roundtable_create (passing the carrier issue_id, topic, and participant count) and keep the returned roundtable_id.

Phase 2 — Collect (parallel Task spawns)

Each participant delivers a one-line stance and a short rationale — the server enforces the caps. As each one responds, and before you synthesize, capture their full position with discussion_append (kind analysis) and then their stance with roundtable_vote. The server flips state → awaiting_human once the Nth distinct non-human vote lands.

Phase 3 — Synthesize

Extract agreements (≥2 endorsements, or unilateral with no opposition) and disagreements (≥2 materially different stances on one question). When there are more than fit, keep the top 4 agreements and top 3 disagreements — one AUQ slot for agreements, up to 3 radio slots for disagreements.

When synthesis yields neither, ask "Retry or skip?"; on skip, call roundtable_close(skip:true, outcome='skipped — no substance') and stop.

Phase 4 — Ratify (one AUQ)

Q1 (multiSelect:true): agreements checkbox. Q2–Q4 (radio): one per disagreement, header ≤12 chars.

Read the full file on GitHub · 55 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 55 lines · 15 tokens per session scan A b41ec380456d

Subscribe to this mod's changes

roundtable is a command published in the GitHub repository trustmybot/plugin (6 stars, last pushed 2d ago), licensed MIT. It adds 15 tokens to every session and 636 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.