New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Round Table

by Soul-Brews-Studio in Soul-Brews-Studio/arra-oracle-skills-cli

Convene a Round Table (โต๊ะกลม): 3–8 souls — persona files, Oracles, or real people's DNA — debate ONE stakes-bearing decision over exactly 4 rounds (OPEN, DEEPEN, CONVERGE, VOTE), one subagent per…

MITAuto-check passedAgent Workflows

Install Round Table

skills CLI
$ npx skills add Soul-Brews-Studio/arra-oracle-skills-cli --skill round-table -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install Soul-Brews-Studio/arra-oracle-skills-cli round-table --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/Soul-Brews-Studio/arra-oracle-skills-cli.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/round-table .claude/skills/round-table && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
round-table
GitHub stars
123
Token cost
~2.1k tokens
SKILL.md length
910 words
Files
7 (incl. scripts)
Skills in repo
31
Repo updated
First seen
Licence
MIT

At a glance

Convene a Round Table (โต๊ะกลม): 3–8 souls — persona files, Oracles, or real people's DNA — debate ONE stakes-bearing decision over exactly 4 rounds (OPEN, DEEPEN, CONVERGE, VOTE), one subagent per…

  • Works in 5 steps: Is this a Round Table question? → Convene: seat the souls (3 to 8) → Rounds: one subagent per soul, all in… → …
  • User says round table
  • SKILL.md covers Files, Step 0 — Is this a Round Table…, Step 1 — Convene: seat the… and Step 2 — Rounds: one subagent…, plus 4 more sections
  • Runs TypeScript scripts from its folder; calls bun and git

What it does

Round Table is an agent skill from Soul-Brews-Studio/arra-oracle-skills-cli. Convene a Round Table (โต๊ะกลม): 3–8 souls — persona files, Oracles, or real people's DNA — debate ONE stakes-bearing decision over exactly 4 rounds (OPEN, DEEPEN, CONVERGE, VOTE), one subagent per soul, then vote with 1–10 confidence; a script does the arithmetic and deadlocks go to the human. Use when user says "round table", "round-table", "โต๊ะกลม", "convene the council", "deliberate this". For "should we …" questions, ask before convening. Do NOT use for brainstorming or one-perspective-at-a-time analysis…

Its SKILL.md is about 2.1k tokens, which your agent loads only when the skill is triggered. The skill folder holds 9 other files, including scripts (for example `examples/adopt-a-pattern.json` and `scripts/table.ts`).

It sits in Agent Workflows, covering Subagents and Brainstorming. The repository describes itself as: Install Oracle skills to Claude Code, OpenCode, Cursor, and 12+ AI coding agents. The licence is MIT.

When your agent uses it

  • User says round table
  • Convene the council
  • Deliberate this
  • One-perspective-at-a-time analysis (use /oracle-prism)

Example prompts

  • “round table”
  • “round-table”
  • “convene the council”
  • “/round-table”

Requirements

  • Node.js

Workflow steps

5 steps, taken from the step headings in SKILL.md.

  1. Is this a Round Table question?
  2. Convene: seat the souls (3 to 8)
  3. Rounds: one subagent per soul, all in parallel
  4. Vote: the script counts, not the model
  5. The minute, and handing it back

What it can do on your machine

Read from SKILL.md and the folder at commit 9feea09. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships 1 file in scripts/ (TypeScript), which the agent can run.

    Shell commands in SKILL.md call:

    • bun
    • git

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use git, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Round Table loads about 2.1k tokens when it runs. Until then it costs about 137 tokens; SKILL.md has 910 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~137
When it runs · the whole SKILL.md, loaded when a task matches
~2.1k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from Soul-Brews-Studio/arra-oracle-skills-cli at commit 9feea09, republished under its MIT licence (© Soul-Brews-Studio). 910 words, ~2,123 tokens.

Download SKILL.mdSave it as .claude/skills/round-table/SKILL.md (or your agent's skills folder). This skill also uses 6 other files; get the full folder from GitHub.
name
round-table
description
Convene a Round Table (โต๊ะกลม): 3–8 souls — persona files, Oracles, or real people's DNA — debate ONE stakes-bearing decision over exactly 4 rounds (OPEN, DEEPEN, CONVERGE, VOTE), one subagent per soul, then vote with 1–10 confidence; a script does the arithmetic and deadlocks go to the human. Use when user says "round table", "round-table", "โต๊ะกลม", "convene the council", "deliberate this". For "should we …" questions, ask before convening. Do NOT use for brainstorming or one-perspective-at-a-time analysis (use /oracle-prism).
argument-hint
"<question>" --agents a,b,c [--options x,y,z] [--aggregator weighted|majority|unanimity-or-escalate]

/round-table — โต๊ะกลม

"Many bodies, one soul" — so one agent can seat many souls at one table, let them argue, and still hand the decision back to the human.

/oracle-prism wears souls one after another, inline. A Round Table seats them together: each soul is its own subagent, hears the others, pushes back, and finally votes with a confidence it has to defend. Bounded by design: 4 rounds, never 5.

Where the shape comes from. A 7-agent, 4-round, confidence-voted meeting that Boy (Pinyo) ran inside his Oracle family. Only its parameters survived — 7 agents, 4 rounds, a confidence vote at the end, reserved for expensive decisions. Mother Oracle rebuilt the round prompts, the 1–10 scale and the aggregation rules from those parameters. So this skill is spiritually faithful, not forensically reproduced: if better prompts turn up, replace prompts/ — don't defend them.

Files

text
round-table/
├── SKILL.md
├── prompts/round1-open.txt … round4-vote.txt   one template per round, {{slots}}
├── scripts/table.ts                            inspect (early stop) + vote (aggregate)
└── examples/adopt-a-pattern.json               worked example — must give 37/50

SKILL_DIR is the base directory printed above this skill.

Step 0 — Is this a Round Table question?

Suggest /oracle-prism instead unless all three hold:

  1. A decision, not a topic — something will be done differently depending on the answer.
  2. Canonical options, 2–4 of them (--options). If the user gave none, propose them and wait for a yes.
  3. Stakes — being wrong is expensive or hard to reverse.

"Should we …" phrasing is not consent to convene. Ask first.

Then find where the minute will live:

bash
# Find oracle root — git toplevel that has CLAUDE.md + ψ/
ORACLE_ROOT=$(git rev-parse --show-toplevel 2>/dev/null)
if [ -n "$ORACLE_ROOT" ] && [ -f "$ORACLE_ROOT/CLAUDE.md" ] && { [ -d "$ORACLE_ROOT/ψ" ] || [ -L "$ORACLE_ROOT/ψ" ]; }; then
  PSI="$ORACLE_ROOT/ψ"
elif [ -f "$(pwd)/CLAUDE.md" ] && { [ -d "$(pwd)/ψ" ] || [ -L "$(pwd)/ψ" ]; }; then
  ORACLE_ROOT="$(pwd)"
  PSI="$ORACLE_ROOT/ψ"
else
  echo "⚠️ Not in oracle repo (no CLAUDE.md + ψ/ at git root). Writing to pwd."
  ORACLE_ROOT="$(pwd)"
  PSI="$ORACLE_ROOT/ψ"
fi
SLUG="$(date +%Y-%m-%d)_<short-slug-of-the-question>"
TABLE="$PSI/lab/round-table"; mkdir -p "$TABLE" "$PSI/inbox/decisions"
MINUTE="$TABLE/$SLUG.md"; STATE="$TABLE/$SLUG.json"

Step 1 — Convene: seat the souls (3 to 8)

Quorum is 3 — two is a debate, one a monologue. Default to 5 when the user names no roster. Every seat needs a soul read from somewhere real — never invented:

Seat kindWhere the soul comes from
A persona fileany path the user gives (resonance_<name>.md, a character sheet, a SKILL.md) — read it whole
An Oracle in your fleetif maw is installed: R=$(maw locate <name> --path) → $R/ψ/memory/resonance/<name>.md, else $R/CLAUDE.md's identity section. Not found → refuse the seat and say so
A real person (multi-DNA)one line — name + the single habit they bring — written or confirmed by the human

Print the roster with each soul's source, the question, the options and the aggregator (default weighted). Gather every fact the souls may use into question_context now — there is no tool use during the rounds.

Create $STATE in the shape of examples/adopt-a-pattern.json (id, question, options, aggregator, agents[] with name, soul, rounds[]) and start $MINUTE with the roster. Flush both after every round, so one flaky subagent never costs the earlier rounds. Persona files may be private — the minute stays in your ψ, not in a post.

Step 2 — Rounds: one subagent per soul, all in parallel

For each round, fill that round's template from $SKILL_DIR/prompts/ and launch one Agent call per soul, all in a single message. Each prompt carries the full soul text, the question, the options and the transcript so far. Subagents debate from soul + context only — no tools.

RoundTemplateEach soul gets
1 OPENround1-open.txtsoul, question, context — hears nobody
2 DEEPENround2-deepen.txt+ every round-1 position; must challenge one by name
3 CONVERGEround3-converge.txt+ rounds 1–2, + 2–4 one-line disagreement axes that you extract
4 VOTEround4-vote.txt+ the full transcript; binding, 5 is forbidden

Parse each reply as JSON. Malformed or off-option → re-prompt that soul once; a second failure empties the seat and the minute says so. Append {round, position, confidence, self_update} to that soul's rounds in $STATE, and the round to $MINUTE.

After rounds 2 and 3, ask the inspector — don't eyeball it:

bash
bun "$SKILL_DIR/scripts/table.ts" inspect "$STATE"
OutputDo
EARLY-STOPeveryone agrees at ≥ 8 — go straight to round 4
SKIP-ROUND-3nobody moved in round 2 — confirm no new evidence arrived, then vote
EXPERT-DEFERENCEone soul at 10, the rest ≤ 3 — ask the human whether to short-circuit
CONTINUErun the next round
Show full SKILL.md (297 more words)Show less

Step 3 — Vote: the script counts, not the model

Write each soul's round-4 answer into final (vote, confidence, rationale) and dissent_note, then:

bash
bun "$SKILL_DIR/scripts/table.ts" vote "$STATE"
ExitMeaningDo
0decidedprint the decision block
2RE-PROMPT — a 5, an off-option vote, a bad numberre-ask only the named souls, vote again
3ESCALATEwrite the deadlock in plain words; the human breaks it
1bad input, or fewer than 3 soulsfix $STATE

What the script enforces:

  • weighted (default) sums confidence per option · majority counts heads · unanimity-or-escalate needs every soul on one side at ≥ 7.
  • Ties break by highest single confidence, then fewer dissent notes, then the human — never an automatic pick.
  • Escalates under every aggregator when the winner holds < 60% of the weighted confidence, when 10s land on opposite sides, or when a dissent cites a value (honesty, safety, consent — override with values_keywords in $STATE). The human decides the hard ones.

Step 4 — The minute, and handing it back

Close $MINUTE with the vote table (soul · vote · confidence · rationale) and the script's block verbatim, e.g.:

text
DECISION: adopt-modified
Weighted confidence: 37/50 (74%) for adopt-modified
Top dissent: none (no opposing votes)
Top hedger: brand (6) — Compromise — would prefer as-is, but unity matters more

Write one line — the decision and the minute's path — to $PSI/inbox/decisions/$SLUG.md, so /recap finds it. Then:

bash
# announce-mode → absolute path (no ψ/, no ~/, no $VAR, no ...).
# Use: echo "marker: $RESOLVED_PATH"  — bash substitutes; never print $VAR literally. See CONVENTIONS.md.
echo "🪑 Minute: $MINUTE"
echo "📥 Decision: $PSI/inbox/decisions/$SLUG.md"

And step out of the table:

  • Every seat was a lens. Write "the finance lens voted…", never "Finance decided…". The souls were worn by the convener; a seat is not that Oracle's or that person's real view. For a real view, ask them (/talk-to if you have a fleet).
  • A decision is advice. The human ships, merges and signs — the table only recommends.

Self-check

bash
bun "$SKILL_DIR/scripts/table.ts" vote "$SKILL_DIR/examples/adopt-a-pattern.json"

Must print DECISION: adopt-modified and 37/50 (74%) and exit 0 — the result the reconstructed spec predicts for its worked example. Anything else means table.ts drifted.

Not in scope

No tool use mid-round · no soul summoning another agent · no human interjection mid-round · no round 5.

© Soul-Brews-Studio, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 6 other files (scripts) in skills/round-table of Soul-Brews-Studio/arra-oracle-skills-cli.

  • SKILL.md
  • examples/adopt-a-pattern.json
  • prompts/round1-open.txt
  • prompts/round2-deepen.txt
  • prompts/round3-converge.txt
  • prompts/round4-vote.txt
  • scripts/table.ts

Open the folder on GitHubat commit 9feea09

Compare with similar skills

Round Table next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Round Table compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Round Table this skillSoul-Brews-Studio/arra-oracle-skills-cli123—~2.1kAutomated safety check: PassMIT
BrainstormAedelon/claude-code-blueprint120—~1.1kAutomated safety check: PassCustom licence
Ideateav/mi102—~2.8kAutomated safety check: PassNone
Lmstudio Subagentssundial-org/awesome-openclaw-skills663—~1.2kAutomated safety check: PassMIT
Subagent DevArman-Kudaibergenov/1c-ai-development-kit166—~551Automated safety check: NotesAGPL-3.0
Write PlanArman-Kudaibergenov/1c-ai-development-kit166—~1.3kAutomated safety check: NotesAGPL-3.0

Similar skills

  • Brainstorm

    Aedelon/claude-code-blueprint

    Multi-agent brainstorming and strategic thinking on any complex topic.

    120 GitHub stars~1.1k tokensUpdated 7 mo ago
    Agent WorkflowsAuto-check passed

  • Ideate

    av/mi

    Timeboxed ideation on a topic using propose-and-critique subagent pairs.

    102 GitHub stars~2.8k tokensUpdated 13 days ago
    Agent WorkflowsAuto-check passed

  • Lmstudio Subagents

    sundial-org/awesome-openclaw-skills

    Reduces token usage from paid providers by offloading work to local LM Studio models.

    663 GitHub stars~1.2k tokensUpdated 7 mo ago
    Agent WorkflowsAuto-check passed

  • Subagent Dev

    Arman-Kudaibergenov/1c-ai-development-kit

    Этот скилл MUST быть вызван когда tasks.md готов и требуется автономное параллельное выполнение субагентами (implementer + reviewer на каждую задачу).

    166 GitHub stars~551 tokensUpdated 4 mo ago
    Agent WorkflowsAuto-check: notes

  • Write Plan

    Arman-Kudaibergenov/1c-ai-development-kit

    Этот скилл MUST быть вызван когда есть design.md и нужно разбить доработку на атомарные задачи (2-5 мин) с точными путями файлов и критериями проверки в tasks.md.

    166 GitHub stars~1.3k tokensUpdated 4 mo ago
    Agent WorkflowsAuto-check: notes

  • Talk

    Abilityai/cornelius

    Conversational partner mode - delegates each exchange to the thinking-partner sub-agent, which embodies the knowledge base as its own memory and engages as an intellectual equal.

    109 GitHub stars~834 tokensUpdated yesterday
    Knowledge ManagementAuto-check passed

More from Soul-Brews-Studio/arra-oracle-skills-cli

All 31 skills in this repo

  • Dig

    Soul-Brews-Studio/arra-oracle-skills-cli

    Mine Claude Code sessions — timeline, gaps, repo attribution, session history.

    123 GitHub stars~1.8k tokensUpdated 6 days ago
    Auto-check passed

  • Forward

    Soul-Brews-Studio/arra-oracle-skills-cli

    Hand off the current session to the next one. An agent skill from Soul-Brews-Studio/arra-oracle-skills-cli.

    123 GitHub stars~2.9k tokensUpdated 6 days ago
    Auto-check passed

  • Project

    Soul-Brews-Studio/arra-oracle-skills-cli

    Clone and track external repos. An agent skill from Soul-Brews-Studio/arra-oracle-skills-cli.

    123 GitHub stars~1.4k tokensUpdated 6 days ago
    Auto-check passed

  • Soul

    Soul-Brews-Studio/arra-oracle-skills-cli

    Bring an Oracle into the current repo — its soul, ALL its skills and its expertise — by link, never by copy.

    123 GitHub stars~1.3k tokensUpdated 6 days ago
    Auto-check passed

  • Trace

    Soul-Brews-Studio/arra-oracle-skills-cli

    Find projects, code, and knowledge across git history, repos, docs, and Oracle.

    123 GitHub stars~2.9k tokensUpdated 6 days ago
    Auto-check passed

  • Recap

    Soul-Brews-Studio/arra-oracle-skills-cli

    Session orientation and awareness — retro summaries, handoffs, git state, focus.

    123 GitHub stars~2.7k tokensUpdated 6 days ago
    Auto-check passed

Categories

Questions about Round Table

What does Round Table do?

Convene a Round Table (โต๊ะกลม): 3–8 souls — persona files, Oracles, or real people's DNA — debate ONE stakes-bearing decision over exactly 4 rounds (OPEN, DEEPEN, CONVERGE, VOTE), one subagent per…. Round Table is an agent skill from Soul-Brews-Studio/arra-oracle-skills-cli. Convene a Round Table (โต๊ะกลม): 3–8 souls — persona files, Oracles, or real people's DNA — debate ONE stakes-bearing decision over exactly 4 rounds (OPEN, DEEPEN, CONVERGE, VOTE), one subagent per soul, then vote with 1–10 confidence; a script does the arithmetic and deadlocks go to the human.

When should I use Round Table?

Round Table fits situations like: user says round table; convene the council; deliberate this; one-perspective-at-a-time analysis (use /oracle-prism).

How do I install Round Table in Claude Code?

Run `npx skills add Soul-Brews-Studio/arra-oracle-skills-cli --skill round-table -a claude-code`. Or copy the skill folder (skills/round-table in Soul-Brews-Studio/arra-oracle-skills-cli) into .claude/skills/round-table in your project. Claude Code loads it when a task matches its description.

How do I install Round Table in Codex?

Run `npx skills add Soul-Brews-Studio/arra-oracle-skills-cli --skill round-table -a codex`. Or copy the skill folder (skills/round-table in Soul-Brews-Studio/arra-oracle-skills-cli) into .agents/skills/round-table in your project. Codex loads it when a task matches its description.

Can I use Round Table in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Soul-Brews-Studio/arra-oracle-skills-cli --skill round-table -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/round-table, .gemini/skills/round-table, .github/skills/round-table and .opencode/skills/round-table in your project.

What does Round Table need to run?

Going by SKILL.md and its folder, Round Table needs TypeScript for the scripts in its folder and the command-line tools its instructions call (bun and git). Our summary lists: Node.js.

Does Round Table access the network?

SKILL.md contains no URLs. Its commands use git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Round Table safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.

What licence does Round Table use?

Round Table is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Round Table use?

About 2.1k tokens (SKILL.md is roughly 8.5k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Round Table?

Skills that share tags, products or a category with Round Table: Brainstorm (Aedelon/claude-code-blueprint, 120 stars), Ideate (av/mi, 102 stars), Lmstudio Subagents (sundial-org/awesome-openclaw-skills, 663 stars) and Subagent Dev (Arman-Kudaibergenov/1c-ai-development-kit, 166 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Round Table?

Soul-Brews-Studio (a GitHub organization) maintains it in Soul-Brews-Studio/arra-oracle-skills-cli, which has 123 GitHub stars. The repository holds 31 skills in this directory. The repository was last updated on October 3, 2026.

Source: Soul-Brews-Studio/arra-oracle-skills-cli on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.