Reads and summarizes shared workspace artifacts left by desks in a workshop bench directory.
npx skills add https://github.com/github/awesome-copilot --skill bench-read
Please help me install this Agent Skill. Skill: bench-read Source: https://github.com/github/awesome-copilot/tree/main/skills/bench-read Command: npx skills add https://github.com/github/awesome-copilot --skill bench-read
In multi-agent workshop setups, desks (specialized agents) leave findings, verdicts, drafts, and reports as files in a shared bench directory. Without a consistent reading process, it is hard to know what other desks produced, what state their work is in, and what needs the operator's attention.
Ask the agent to show what is currently on the bench and summarize what each desk has produced so far.
High-intent search queries and answers about installing and using this skill.
Read the bench directory at the workshop root, which is the shared cross-desk workspace. Start with the directory listing, then read each desk's journal to understand what was done before opening specific artifacts.
The bench is the bench/ directory at the workshop root, established as the shared workspace for cross-desk work. Desks leave findings, verdicts, drafts, and reports there as files rather than sending messages to each other.
Read journals first. Each desk's journal records what it worked on and where it left things, so the most recent entry tells you the current state. Then read the specific artifacts the journal points to.
No. Another desk's verdict is input, not authority. Desks have equal standing, so you should read the underlying artifact and form your own independent assessment before acting on it.
Shared cross-desk artifacts belong in bench/ at the workshop root. Desk-local work lives under desks/<desk-name>/, including that desk's journal and artifacts still local to it. Check desk-local directories only when you need a specific desk's own work.
garden-coRun structured cronometro benchmarks for Jazz packages with worker-thread isolation.
hungv47Review completed changes with layered evaluation and generate dated artifact reports.
tnezCopies the latest Claude session JSONL log to the local PARA/DESK/outbox.
TitoPrauseeCoordinate asynchronous AI agent messages through GitLab issues.
