Skills/npm/codex-claude-council

codex-claude-council@0.21.1

CAUTION· startup only

Publisher claims: "Ask OpenAI Codex (ChatGPT) and Claude the same question through your own signed-in CLIs; they answer independently, then reply to each other until they agree on one answer. MCP server + plugin for Claude Code, Claude Cowork and Codex." Observed in sandbox: no network, touches nothing outside the project, writes no memory.

Registry npm (MCP server)Publisher hamza-aziz-ai · publishing for 33 daysInstalls 5,638Last detonated 3d ago
What would fix thisWatchDispute verdict

Coverage

Startup only; authenticated tool behavior not tested.
Tested
server startup over stdio (initialize sent) · every file read, file written and host contacted by what ran
Not tested
the tool list (the server exited before answering) · what the tools do when called with real credentials
Needs
COUNCIL_TOKEN, TUNNEL_TOKEN — give it to the server where it runs, through vet-skills run --secrets, so the value never enters the server's environment. Never paste a credential into a chat, a ticket or a report.

What this means for the label: Caution here covers what ran. The server's tools were not exercised with a real credential, so nothing about their behavior is asserted; the credential summary below says what to do with that.

Before you give it your token

It will ask for
ANTHROPIC_API_KEYCOUNCIL_DEVICE_AUTHCOUNCIL_TOKENMAX_TOKENSOPENAI_API_KEYTUNNEL_TOKEN or --with-api-key / --device-authConfiguration, not credentials: CLAUDE_CONFIG_DIR, CODEX_HOME, ComSpec, COUNCIL_BANNER, COUNCIL_CONFIG, COUNCIL_PUBLIC_URL, KONSOLE_VERSION, PATHEXT, TERM_PROGRAM, WT_SESSION
Where it talks to
no vendor host named in the codeAlso githubusercontent.com, cloudflare.com, mermaid.ink, claude.com, claude.ai — not the publisher's domain; check what for before connecting.
What the honeypot saw
no credential files read · no network
What we could not test
the tool list (the server exited before answering) · what the tools do when called with real credentials

Connect it with a scoped credential, under the guard, and read the third-party hosts first. The code names githubusercontent.com, cloudflare.com, mermaid.ink, claude.com, claude.ai as well as its own vendor. The guard injects ANTHROPIC_API_KEY, COUNCIL_DEVICE_AUTH, COUNCIL_TOKEN, MAX_TOKENS, OPENAI_API_KEY, TUNNEL_TOKEN only into requests to the host you bind (and its subdomains) and refuses the rest.

$vet-skills run --net declared -- npx -y codex-claude-council

Never paste a credential into a chat, a ticket or a report, including ours. How the guard holds it.

Is [email protected] safe to install?

KillerSkills ran it in the KillerSkills honeypot sandbox (2 scenarios; startup only; authenticated tool behavior not tested) on 2026-11-11 and scored it 68/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Tunneling service URL; Pipe-to-shell pattern; Access to sensitive file paths. Summary: no network, touches nothing outside the project, writes no memory.

Behavior trace

Honeypot run · 46 s · 2 scenarios · 0 deviations
t+CategoryObserved action
0.0sprocessno install scripts in bundle
5.9sprocessdependency install exit 0
26sprocessrelaunched with a decoy COUNCIL_TOKEN, TUNNEL_TOKEN (tagged; nothing real)
46sclaimedstartup only: exited before answering, asking for COUNCIL_TOKEN, TUNNEL_TOKEN (a decoy value was refused)
46sprocessscript backend runs setup only
46sprocessno install scripts to re-run

Static findings

HIGH
Tunneling service URLscripts/cli.mjs:183 · Tunneling service URL — may route traffic to attacker infrastructure* cloudflared reports its https://<words>.trycloudflare.com address.Fix: Replace encoded or assembled strings with plain literals and remove undeclared network calls.
AST07
CRITICAL
Pipe-to-shell patternsrc/install.mjs:24 · Pipe-to-shell pattern — downloads and immediately executes remote code' macOS / Linux / WSL: curl -fsSL https://claude.ai/install.sh | bash',Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST06
MEDIUM
npx invocation of non-well-known packagesrc/api.mjs:27 · npx invocation of non-well-known package — may execute untrusted codethrow new Error(`the API mode needs the providers' SDKs, which this copy of ${NAME} doesn't include. Run \`npx -y ${NAME} api-setup\` once, then try again.`);Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packageskills/codex-claude-council/SKILL.md:68 · npx invocation of non-well-known package — may execute untrusted code- For a missing CLI, a sign-in problem or a usage limit, tell the user plainly and suggest `npx -y codex-claude-council doctor`. Do not retry in a loop.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
LOW
Prompt injectionREADME.md:248 · Prompt injection — known jailbreak technique reference · informational: not an instruction file| **ChatGPT** | Turn on developer mode (**Settings → Apps → Advanced settings**), then create a connector with the URL |Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
HIGH
Access to sensitive file pathssrc/config.mjs:73 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files// COUNCIL_CODEX_AUTH / COUNCIL_CLAUDE_AUTH: the same, for a setup configured by environment (Docker's .env).Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
npx invocation of non-well-known packagesrc/install.mjs:267 · npx invocation of non-well-known package — may execute untrusted code`${c.gold('4.')} Optional defaults (model, effort, timeout): ${c.link(`npx -y ${NAME} config --init`)}, then edit ${c.link(userConfigPath())}`,Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagesrc/skills.mjs:108 · npx invocation of non-well-known package — may execute untrusted code+ 'Install one with: npx -y codex-claude-council skill add <GitHub URL or SKILL.md path>');Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
HIGH
Hidden zero-width characters detectedsrc/config.mjs:39 · Hidden zero-width characters detected — may conceal contentreturn JSON.parse(text.replace(/^/, ''));Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST07
MEDIUM
Process kill commandsrc/process.mjs:93 · Process kill command — agent can terminate processes/** Kill a process and its children; resolves once the kill has been carried out (taskkill is a separate process). */Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
npx invocation of non-well-known packagesrc/share.mjs:98 · npx invocation of non-well-known package — may execute untrusted code<code>npx -y codex-claude-council install</code>Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagesrc/install.mjs:292 · npx invocation of non-well-known package — may execute untrusted codeconst notInstalled = error => new Error(`${error.message} If ${NAME} is not installed yet, run \`npx -y ${NAME} install\` instead.`);Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
LOW
Authorization escalation languageREADME.md:248 · Authorization escalation language · informational: not an instruction file| **ChatGPT** | Turn on developer mode (**Settings → Apps → Advanced settings**), then create a connector with the URL |Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
HIGH
Host or code assembled from base64 fragmentsscripts/npm-readme.mjs:26 · Host or code assembled from base64 fragmentsreturn `[![Diagram ${diagram}](https://mermaid.ink/img/${Buffer.from(code).toString('base64url')})](https://github.com/${REPO}/blob/${ref}/README.md)`;Fix: Put hosts, commands and strings in plain source. Decoding at runtime looks identical to malware and costs points on its own.
AST07
HIGH
Hidden zero-width characters detectedsrc/install.mjs:156 · Hidden zero-width characters detected — may conceal contentconst config = exists ? JSON.parse(readFileSync(path, 'utf8').replace(/^/, '') || '{}') : {};Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST07
MEDIUM
Remote installer piped to shellsrc/install.mjs:24 · Remote installer piped to shell' macOS / Linux / WSL: curl -fsSL https://claude.ai/install.sh | bash',Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST07

Cost to the agent

Instructions loaded
≈2,337 tok
1 file · 29 in bundle
Tool list per prompt
≈0 tok
0 tools · 0 per tool
Handshake
no reply
launch to tools/list, stdio
Undocumented tools
0
description under 20 chars · avg 0

Tokens estimated at four characters each. The tool list is re-read by the agent on every turn, so it is the number to compare between servers that do the same job: ranked across the index.

Known vulnerabilities

No advisories in OSV for codex-claude-council@0.21.1 as of 3d ago. This is the package itself, not its dependencies.

What would make this Verified

12 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%
  1. Pipe-to-shell pattern≈ +9 on the totalBuild commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
  2. Unclaimed publisher≈ +8 on the totalClaim hamza-aziz-ai: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
  3. Obfuscated code≈ +5 on the totalRemove base64-assembled strings, eval of decoded text and minified blobs. Reviewers and scanners both treat unreadable code as hostile.
  4. Tunneling service URL≈ +4 on the totalReplace encoded or assembled strings with plain literals and remove undeclared network calls.
  5. Access to sensitive file paths≈ +4 on the totalStop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
  6. Hidden zero-width characters detected≈ +4 on the totalKeep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
Publish a new version to npm, then queue it here. Not your skill? Claim the handle to get verdict changes by email.Claim hamza-aziz-aiPropose a patch to the maintainer