What is the-council CLAUDE.md?

Project instructions for a skill that asks OpenAI Codex and Google Gemini for separate advice, then combines their analyses into one recommendation. Claude Code is the main agent coordinating this process.

Use it to run preflight checks, synchronize project context, and consult Codex and Gemini in parallel through shell scripts.

It gives a coding agent more than one independent view when reviewing code, planning architecture, debugging, or answering a general question.

How DantesPeak85/the-council configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to DantesPeak85/the-council. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/DantesPeak85/the-council/main/CLAUDE.md

Clone the repo

git clone --depth 1 https://github.com/DantesPeak85/the-council

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/instructions/dantespeak85/the-council/claude-md/github.svg)](https://agentmods.dev/instructions/dantespeak85/the-council/claude-md)

Your own site

<a href="https://agentmods.dev/instructions/dantespeak85/the-council/claude-md"><img src="https://agentmods.dev/badge/instructions/dantespeak85/the-council/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/instructions/dantespeak85/the-council/claude-md"><img src="https://agentmods.dev/badge/instructions/dantespeak85/the-council/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 1,595 This file is loaded in full into every session.

When invoked 1,595 The same file — it is already loaded in full.

Security scan D 2 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin 97% copy Near-identical to another mod 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.

Model	Per session	Once invoked
Fable 5.1	$0.01595	$0.01595
Opus 5.5	$0.00638	$0.00638
Sonnet 5.5	$0.00319	$0.00319
Haiku 4.5	$0.00160	$0.00160

Measured 1mo ago against content hash 06951f0e48e6, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade D, and why

the-council CLAUDE.md scanned grade D with 2 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 1mo ago.

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.

Reads agent configuration directoriesmediumAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

| `CODEX_MODEL` | unset → `~/.codex/config.toml` | Codex advisor model (passed via `-m`) |

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

Invocation creates `.council-tmp/council_<mode>_YYYYMMDD_HHMMSS/` (where `<mode>` is `full`, `codex`, or `gemini` depending on which advisors ran) in the working directory, containing `prompt_final.txt`, `review_request.

Origin

This is a copy

97% identical to the-council AGENTS.md — 8 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.

CLAUDE.md · 59 lines

How it starts

The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

The Council is a Claude Code Skill plugin that convenes OpenAI Codex (codex CLI) and Google Gemini (agy CLI, Google's Antigravity client running Gemini 3.5 Flash by default) as an advisory board. Both run in parallel as effectively read-only consultants, returning independent analyses that Claude synthesizes into unified recommendations.

This is not a Node/TypeScript application — it's a pure Bash + Markdown skill with no build system, package manager, or test framework.

Architecture

skills/the-council/
├── SKILL.md                        # Operational guide (skill definition + workflow)
├── references/prompt-templates.md  # 4 prompt templates (review, architecture, debug, general)
└── scripts/
    ├── council_preflight.sh        # Detects available CLIs + auth status (cached 2h)
    ├── council_sync.sh             # Copies CLAUDE.md → AGENTS.md (Codex only)
    └── council_invoke.sh           # Parallel CLI invocation: timeouts, OS sandbox, snapshot safety net, response validation

Execution flow: Preflight → Sync context → Compose prompt from template → Invoke advisors in parallel → Synthesize responses (consensus/divergence/recommendation) → Generalize learnings into CLAUDE.md/AGENTS.md.

Operating modes are determined by preflight: Full Council (both CLIs), Codex-only, Gemini-only, or abort (neither available).

Key Design Decisions

Read-only enforcement (1.3.0+): Both advisors run with OS-level deny-write sandboxes on macOS. Codex uses its built-in --sandbox read-only. Gemini (agy) is wrapped in sandbox-exec with a deny-write profile at skills/the-council/scripts/council_sandbox.sb; writes are allowed only to ~/.gemini/, the per-invocation .council-tmp/<...>/ dir, system temp, and agy-specific ~/Library/Caches/ subdirs. On non-macOS, the script refuses to run agy unless --allow-unsandboxed-gemini is passed.
Diff safety net (1.3.0+): pre/post-invocation snapshot of $WORK_DIR (git: HEAD + status with --ignored=traditional + sha256 of tracked + untracked + gitignored files; non-git: find + sha256). Snapshot excludes .council-tmp/ (response files) and .antigravitycli/ (agy's housekeeping dir). Any other change fails the invocation with exit 2 and a [COUNCIL_SAFETY_NET] banner. See SKILL.md for the historical 1.2.x ghost-write incident this replaces.
Gemini invocation (1.4.0): Gemini reviews the INLINED prompt only — no project --add-dir, no repo exploration. The composed prompt is written into the per-invocation .council-tmp/<run>/ dir, which becomes the advisor's only workspace. The backend is resolved at runtime: gemini CLI (single-shot, stdin, JSON envelope) when a paid GEMINI_API_KEY/GOOGLE_API_KEY is set and gemini is on PATH, else agy (pty-wrapped via script, sandboxed). The old maxSessionTurns settings injection was a verified no-op and is gone — ~/.gemini/antigravity-cli/settings.json is never touched.
Prompt sanitization: jsr: import strings are redacted from the Gemini prompt before invocation to avoid spurious safety/policy blocks from agy.
Empty-response retry: If agy exits 0 with no stdout and no stderr, the invocation is retried once before being reported as failed.
Response validation (1.4.0): an advisor fails only on empty output, a script-written [COUNCIL-ADVISOR-FAILURE] placeholder, a nonzero exit with no verdict line, a refusal, or non-engagement (a short verdict-less reply to a large prompt). The old response-body word-grep (policy/blocked/quota/rate limit) is gone — it false-failed substantive reviews; stderr noise now demotes to an advisory *_warnings.log, not a failure.
Parallel execution: Both advisors run as background processes with configurable timeout (COUNCIL_TIMEOUT, default 600s).
NVM compatibility: council_invoke.sh loads NVM environment to find globally-installed CLI tools.
macOS support: Uses gtimeout (from coreutils) when available, falls back to timeout or no timeout.
Graceful degradation: Works with one advisor if the other is unavailable; shows install help if neither exists.

Read the full file on GitHub · 59 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.

1mo ago Changed · -162 tokens per session scan D → B 24b2a5a7b775
2mo ago First seen · 59 lines · 1,595 tokens per session scan D 06951f0e48e6

Subscribe to this mod's changes

the-council CLAUDE.md is an instructions file published in the GitHub repository DantesPeak85/the-council (3 stars, last pushed 15d ago), licensed MIT. It adds 1,595 tokens to every session, about $0.0064 per session on Opus 5.5. A static security scan graded it D with 2 findings (reads agent configuration directories, recursive force delete). It is 97% identical to the-council AGENTS.md, differing in 8 lines, and is treated as a copy.

Building agents Claude Code Code review agent-skill ai-advisory multi-model skill

vscode buildNext.instructions.md

Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).

microsoft/vscode · 9,715 tokens

the-council: Instructions file for Claude Code