Roundtable
PassAudited by ClawScan on May 1, 2026.
Overview
Roundtable is a coherent instruction-only multi-agent reasoning skill, with disclosed costs, sub-agent sharing, and optional session logging that users should notice.
This skill appears safe to install if you want multi-agent debate-style answers. Before using it, be aware that full mode can make several model calls and that optional logging may save your questions and council outputs under memory/roundtable.
Findings (3)
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.
Using the skill may consume more model calls, time, and budget than a normal single-agent answer.
The skill intentionally dispatches multiple specialist model runs, and Round 2 can double the number of model invocations.
Spawn 3 specialized sub-agents in parallel to tackle complex problems... then optional Round 2 cross-examination
Use it for complex questions, and consider quick mode, budget settings, or confirmation before dispatch for cost-sensitive tasks.
If logging is enabled, sensitive questions or outputs may remain in memory and could influence future use.
The skill can persist council sessions for later reference, which may include user questions and generated analyses.
Should I save council sessions for future reference? A) ✅ Yes, save to memory/roundtable/ B) ❌ No logging
Disable logging for sensitive topics, or periodically review and clear saved roundtable logs.
Information from the original question and intermediate analyses is intentionally passed among sub-agents.
The workflow shares intermediate agent outputs between agents for cross-examination; the artifacts acknowledge the contamination risk and describe a mitigation.
Round 1 findings reused in Round 2 are treated as potentially contaminated and cross-verified
Avoid including secrets in prompts, and review final answers for errors that may have propagated through the council.