Show the reader an interactive chess board in the agent-repl feed for a game loaded in a CEE CLI session — the CEE CLI webapp's own widget, with its move list, engine lines and per-square engine answers. Finds the session and its live game id, then calls agent-repl's showchessboard tool. Only CEE CLI sessions can be…
Answer a question about agent-repl, the Emacs-fronted Claude workspace system in this repository — what it is, how it works, what a feature does and how to use it, and where its user guide, architecture and developer documentation, protobuf contract, and code live. Read-only, never implements or edits anything without…
Investigate the agent-repl system across Emacs, daemon, shim, webapp, sidecar, store, state, deployment, logging, and test coverage. Use for current or historical workspace failures, wrong colors or stuck states, missing conversation data, service-health questions, source-versus-running readiness…
Set up everything agent-repl needs on a computer — check every requirement, install what is missing or outdated (Emacs with xwidgets, toolchains, terminal-notifier, the store, sidecar and metrics-store services, skills, the agent-repld command on PATH, the git hook), walk the user through what only they can provide…
Land work through the agent-repl merge queue, the ONLY way into master in this repository, and control the queue. A merge is one of two kinds. Finishing THIS WORKSPACE (workspace finished here) deploys its own branch as its repository's .agent-repl/deploy.yaml says (for agent-repl: landed where its topology says, then…
Read agent-repl's always-on runtime metrics — user-facing latencies (workspace switches, page boots, vendor first output), how long restarts take (deploys and their phases, daemon boots, store and sidecar restarts, each workspace's session start, shim replacements, handover transfers), CPU, memory, disk, energy and…
Profile a live agent-repl across every runtime at once — CPU and memory of the daemon, store, sidecar, every shim, Emacs and the webapp's pages and WebKit processes — and rank the user-visible latencies (workspace open, switch, first switch to a never-shown workspace, page load, prompt echo) against how perceptible…
Operate the agent-repl runtime — build it, bounce its backends, hard bounce everything including Emacs, roll a deploy out, hot-reload elisp into the running Emacs, check health and readiness, read logs, hold or release a repository's deploys, prune agent-repl's stored data, evaluate elisp in the running Emacs and read…
Write or change agent-repl tests (Go unit/integration/e2e, vitest, ERT, shell harnesses) so they are fast and cheap WITHOUT losing any coverage, and so every new or changed test obeys the host CPU governor. Use whenever an agent adds, changes, speeds up, or stabilizes a test anywhere under modules/app/agent-repl…
Work with agent-repl workspaces through the agent API: create a workspace (a child of this one or of another, or a primary workspace off the main branch, with a fresh context or one forked from a workspace''s conversation, as a one-shot or not), prompt another workspace, read a workspace''s status, list them all, or…
Change the agent harness agent-repl gives its agents — the metaprompt every session loads, a repository's AGENTS.md or CLAUDE.md, a skill, the daemon's prompt briefs, or any agent convention — by putting each instruction in its ONE right home, removing every restatement of it elsewhere, keeping a skill's availability…
Renew an expired or missing GNS token yourself, without asking, signing in ONLY through the work account's own Chrome profile. Use whenever a gns command or the LLM gateway fails on authentication (for example HTTP 401: Invalid or expired JWT, "not logged in", "token expired"), whenever a GNS token is missing, and…
★not rated 3 yesterdayA0 tokens
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: