Council logo

Council

Community
chrisblattman
council

Develop or review a plan, draft, workflow, decision, or argument using independent council members and a fresh synthesis. Use for an explicit council request, including dictated counsel. Discussing councils alone is not an invocation.

Overview

Publisherchrisblattman
Repositoryclaudeblattman
Skill namecouncil
Stars
468
Forks
79
Bundled files
10
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 10 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by chrisblattman on GitHub. Read the source before you install it.

Installation

Install the Council AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/chrisblattman/claudeblattman/tree/main/codex-starter/skills/council
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/chrisblattman/claudeblattman.git /tmp/claudeblattman
mkdir -p .claude/skills
cp -r /tmp/claudeblattman/codex-starter/skills/council .claude/skills/council
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Council in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Council on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Council is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Council

Read references/host.md, references/policy.json, and references/synthesis.md next to this skill. These files and this workflow are generated from one source for both apps. Do not invoke a separate prompt formatter or the Ask Agents conversational workflow.

Select the run

The starter default, explicitly chosen for beginners, is three local members: skeptic, pre-mortem, and completeness checker. Read their complete definitions in references/personas/. Label this a local custom council, not Chris's provider-balanced Light mode. Use three fresh workers in the current app; the orchestrating conversation is not a council member.

Honor an explicit provider/count/lens request; never reinterpret it as the default. The archive helper supports the three default local members, plus Chief of Staff as an optional fourth member only when explicitly requested. Never add or substitute Chief of Staff automatically, even for decisions. Pass --chief-of-staff to plan when the user asks to include it. If the requested custom panel differs, explain that limitation and offer a supported run only with the user's agreement; do not silently add services, change counts, or call it balanced. Clarify only material ambiguity. Do not require an existing draft: members can independently develop an answer or propose an approach against the same brief.

Chris's explicit Light means one Claude, one Codex, and one Gemini member, followed by fresh synthesis. Full means two per provider, then a fresh adversary, then fresh synthesis. Different providers can run concurrently; two Full members from one provider run sequentially. These counts, lenses, and chair rules are preserved in references/policy.json.

Current runtime boundary: this starter ships the local native-worker route. Chris's audited balanced runtime (including Gemini through Antigravity, model-pinned execution, and bound chair receipts) has not been ported into this starter kit. If Light/Full is requested, explain this specific missing dependency before launching anything. Offer the supported local council or an explicitly labeled custom outside opinion only if the user chooses it. Do not feed a starter manifest into Chris's incompatible resolver/runner, use Gemini CLI as its replacement, or claim that installing Ask Agents enables audited Light/Full. See references/runtime-status.md.

Prepare a neutral brief and durable archive

Preserve the user's question verbatim; add a neutral evidence index, governing constraints, requested output, and stated success criteria. Do not insert the host's preferred answer, invent requirements, or rewrite an ambiguous objective into a different task. Ask about a missing fact only when it materially changes the work. Stable source files remain at original paths; do not recursively copy source folders. Declare intentionally mutable files or directory pointers as live evidence.

Write the brief to a temporary file, then use the bundled helper (Python 3):

text
python3 <skill-directory>/scripts/council.py plan --host <claude|codex> --mode local --brief-file <brief.md> --output <project>/council/YYYY-MM-DD_topic --evidence <stable-file>

Repeat --evidence as needed; use --live-evidence for intentionally mutable sources or directory pointers. The output directory must be new; use a fresh suffix for another run on the same date. The helper freezes PROMPT.md, MANIFEST.json, MANIFEST.md, exact seat-prompts/, and the evidence identities. It creates private raw/ and reports/ directories. Keep these project records out of public releases.

Independent work

Dispatch each generated seat prompt unchanged through the host's native subagent mechanism in references/host.md. Give each worker the shared task and evidence, complete assigned lens, and its unique report path. Do not give members earlier answers or other members' reports. Each worker can choose its research depth and useful specialist checks; discourage recreating the entire panel. Do not add arbitrary turn, time, spending, token, or output ceilings. Honor the host's actual permissions and user restrictions; never weaken them to mimic Chris's machine.

Members analyze and report; they do not revise source artifacts or shared control files. Wait for actual completion. Save each full answer exactly as returned, including partial substantive content and separate additional deliveries. Preserve execution identifiers and observable model facts without inventing telemetry. Record each terminal outcome:

text
python3 <skill-directory>/scripts/council.py record --run <archive> --seat skeptic --state succeeded --report-file <exact-report.md> --execution-id <actual-worker-id>

Use the other seat IDs printed by plan; failed, blocked, and partial are also supported. Include --note for execution/access caveats and repeat --addendum-file for separate additional deliveries. The helper copies the full report without overwriting an earlier delivery and checks stable evidence for changes. A partial report or failed execution remains available to synthesis, not a successful vote. Do not rerun merely to obtain cleaner formatting. An unavailable native subagent facility is a blocker; never impersonate three members in the host conversation.

Challenge and fresh synthesis

For the default local run, go directly to synthesis. An explicit request for a dedicated challenge may add a fresh adversary after the independent reports, before synthesis; it remains a custom run. Request it with plan --adversarial. This follows the audited ordering and does not make a local panel Full. Do not launch an adversary when independent reports are incomplete or their evidence changed; disclose the omission and synthesize available reports with limitations.

text
python3 <skill-directory>/scripts/council.py packet --run <archive> --stage adversary
python3 <skill-directory>/scripts/council.py record --run <archive> --stage adversary --state succeeded --report-file <exact-report.md> --execution-id <actual-worker-id>
python3 <skill-directory>/scripts/council.py packet --run <archive> --stage synthesis

Use each exact frozen STAGE_PROMPT.md with a new worker. Packets use anonymous report labels, preserve distinct deliveries as one member's contribution, and include source/verification caveats. Never edit raw reports to force agreement. The adversary checks shared assumptions, source errors, missing alternatives, minority major findings, and unsupported compromises; it does not vote or synthesize. Record an adversarial failure once; do not keep retrying it.

The synthesizer compares reasons and evidence, distinguishes agreement from truth, preserves major minority findings, and labels its own inferences. It produces the requested useful result or recommendation with unresolved questions and decisions for the user. Follow references/synthesis.md. The host does not perform this stage inline. Record synthesis through the same helper, which writes SYNTHESIS.md. A failed primary permits the one predeclared fresh fallback in the manifest; record both attempts. If both fail, return clearly labeled unsynthesized reports without a council verdict. Do not silently turn a failed external provider into the host's answer.

text
python3 <skill-directory>/scripts/council.py status --run <archive>

Show the actual completed/failed members, source limitations, challenge status, synthesis result, and durable synthesis link. The portable helper verifies file integrity, not independently observed provider identity or every source read. Never describe it as Chris's certified balanced execution.

Revision and another run

Let the user accept, reject, or question findings. Revise the actual artifact only when authorized, preserving the prior version. A requested further review starts a new archive and fresh members with the revised artifact, original purpose, evidence, and updated user decisions. Independent members do not inherit earlier debate; after their reports, the synthesis may check whether earlier material objections were resolved. Do not start an automatic multi-round debate or keep looping until agreement. Saving recurring preferences requires the user's direction; it is not an automatic consequence of review. /done saves the handoff when requested.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Council AI skill do?

Develop or review a plan, draft, workflow, decision, or argument using independent council members and a fresh synthesis. Use for an explicit council request, including dictated counsel. Discussing councils alone is not an invocation.

Why use Council on TypingMind?

Because you install it once and use it with any model. Council is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Council in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/chrisblattman/claudeblattman/tree/main/codex-starter/skills/council. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Council?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Council?

As many as you like. As long as a model supports skills, you can use Council with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Council AI skill free?

Yes. It is published on GitHub by chrisblattman under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇