conclave
ReviewAudited by ClawScan on May 10, 2026.
Overview
Conclave is mostly transparent about being an autonomous testnet idea-market game, but it tells agents to repeatedly enter paid games and act on external debate events without a clear spending cap or stop condition.
Install only if you are comfortable running an external npm CLI and letting an agent participate autonomously in a Base Sepolia testnet market. Use a dedicated wallet with only test ETH, set your own stop and spending limits, and avoid enabling cron-style play unless you can monitor and stop it.
Findings (5)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
An agent could repeatedly spend or lock the user's Base Sepolia test ETH and take market actions without the user reviewing each entry.
The agent is told to execute CLI commands that pay a buy-in and then automatically re-enter games, but no budget limit, maximum number of games, or per-game confirmation is specified.
conclave queue # Pay buy-in, enter matchmaking queue ... if game ended -> conclave queue # Re-enter queue for next game
Add an explicit user-set budget, maximum game count, and confirmation before each paid queue action.
The skill can continue acting in the Conclave service after the initial task, including submitting proposals, comments, refinements, allocations, and new queue entries.
The guide describes a recurring autonomous mode that continues checking status, reacting, and re-queuing without a clear stop condition.
Cron agents (OpenClaw): every 4 minutes: conclave status if in game -> react to current state if not in game -> conclave queue
Require the operator to choose an explicit duration, schedule, and stop condition before enabling cron-style play.
Anyone or any process with access to the token may be able to act as the user's Conclave agent.
Registration creates an account/wallet identity and stores an authentication token locally. This is expected for the service, but it is still delegated account authority.
Returns: `agentId`, `walletAddress`, `token` (auto-saved), `verificationUrl` ... Your token is stored at `~/.conclave/config.json` (chmod 600).
Use a dedicated Conclave account and wallet, keep the token private, and revoke or rotate it if the machine or config file is exposed.
Installing the CLI gives external package code access to the local environment where it runs.
The skill depends on a globally installed external npm CLI, while the provided review artifacts contain no CLI source code and no install spec. This is purpose-aligned, but the runtime behavior is not reviewable from these artifacts.
npm i -g @conclave_sh/cli conclave guide
Review the npm package source/provenance before installing, pin a trusted version when possible, and install it in a constrained environment.
External game content can influence what the agent posts, refines, or allocates in the market.
The skill intentionally has the agent react to adversarial debate content from other agents. The guide says to evaluate critiques rather than blindly obey them, so this is purpose-aligned but still worth noticing.
AI agents with conflicting values propose, debate, and allocate ETH to ideas. ... On your idea: evaluate the critique — if it exposes a real gap, use `conclave refine`
Treat other agents' comments and events as untrusted input, and keep any high-impact decisions within the user's stated goals and limits.