Roundtable
v0.4.1Multi-agent debate council — spawns 3 specialized sub-agents in parallel (Scholar, Engineer, Muse) for Round 1, then optional Round 2 cross-examination to ch...
⭐ 2· 909·8 current·8 all-time
byRobby@robbyczgw-cla
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (multi-agent debate council) match the SKILL.md: it spawns three specialist perspectives, supports config/presets/templates, optional Round 2 cross-examination, and session logging. There are no unexpected credentials, binaries, or config paths required that would be unrelated to the stated purpose.
Instruction Scope
SKILL.md is the runtime instruction set and stays largely within scope: parsing user commands, running three specialist roles in parallel, optionally running a Round 2, prompting an interactive setup, and writing/reading a config.json in the skill workspace. Two items to note: 1) it logs sessions to a fixed workspace path (memory/roundtable), which means user queries and agent outputs may be persisted; 2) it references web results for Scholar and asks specialists to treat web results as untrusted, but it does not specify particular remote endpoints or require external credentials. These are reasonable for the feature but worth informing the user.
Install Mechanism
This is an instruction-only skill with no install spec and no code to download or execute. The package.json, README, and config.example.json are documentation-only artifacts; nothing in the manifest pulls external archives or installs binaries.
Credentials
No environment variables, credentials, or privileged config paths are requested. The only persistent artefacts are a config.json written to the skill directory after explicit user confirmation and optional session logs in memory/roundtable. Those are proportional to the feature (config and auditable logs).
Persistence & Privilege
The skill does not request always:true and is user-invocable; it can invoke multiple internal sub-agents (normal for a multi-agent skill). It will write config.json (after explicit confirmation) and may write session logs to memory/roundtable. Persisting full session logs could expose sensitive user inputs — the SKILL.md documents this and enforces a fixed safe path, but users should be aware of stored content and retention.
Assessment
This skill appears coherent with its stated purpose. Before installing, consider: 1) Cost: by design Roundtable runs 3 specialists (quick) or 6 runs (full), so expect ~3x–6x token/cost amplification versus a single-agent response; use the --quick, --budget, and --confirm flags if cost matters. 2) Persistence: if you enable logging the council will save sessions to memory/roundtable in the workspace (config.json is written only after you confirm). Those logs can contain the full user query and agent outputs — do not enable logging if you expect to process highly sensitive secrets. 3) Autonomy: the skill spawns multiple sub-agents automatically when invoked; that is intended behaviour but increases blast radius of any prompt mistakes. If you need more assurance, run /roundtable setup, review the generated config.json, disable logging, and use conservative model/preset choices. Overall there are no hidden endpoints or credential demands, but be mindful of cost and stored session data.Like a lobster shell, security has layers — review code before you run it.
latestvk9775p52bbthcefk8xmz9kdxw9827t4r
License
MIT-0
Free to use, modify, and redistribute. No attribution required.