open-cowork

User-facing Gateway assistant for OpenCode, Telegram, and WhatsApp sessions. Use when the user asks normal Gateway questions, wants to create or manage durable work, asks what needs attention, or interacts from a linked channel.

Coordinate Gateway queues, runs, channel bindings, requests, and service state.

Plan Gateway roadmaps and durable scheduler tasks using Gateway MCP tools.

Spec-driven review and verification gate for Gateway tasks across code, docs, slides, research, operations, and other deliverables.

Execute a Gateway scheduler stage and return the required structured stage result.

Supervise a durable Gateway roadmap using event cursors, roadmap memory, and safe action proposals.

Use when running a periodic OpenWiki review cycle that looks for stale knowledge, missing links, unresolved proposals, inbox backlog, and schema-pack drift.

Use when proposing OpenWiki page edits, source additions, or synthesis pages through proposal-safe configured OpenWiki MCP tools.

Use when triaging OpenWiki inbox items, transcripts, meeting notes, or user-submitted knowledge into source records, page proposals, and reviewable wiki changes.

Use when turning transcript evidence into OpenWiki meeting pages, entity updates, decisions, action items, and proposal-safe wiki changes.

Use when working with OpenWiki or configured OpenWiki MCP tools to search, add sources, propose pages or edits, validate proposals, review/apply changes, monitor agent activity, or explain wiki Git history.

Use when editing or proposing OpenWiki content where Spaces, visibility, policy scopes, or private records may affect what the agent can read or disclose.

Use when drafting reviewable OpenWiki proposals from retrieved evidence, link suggestions, schema-pack requirements, or inbox-derived material.

Use when answering questions from OpenWiki records with configured OpenWiki MCP search, reads, claims, source tracing, and cited IDs.

Use when handling meeting transcripts, call notes, or other transcript-like OpenWiki inbox items as untrusted evidence.

Design and author custom OpenCode agents for Open Cowork, including routing, skills, tools, and permission boundaries.

Run an auto research / autoresearch loop for measurable improvements. Use when optimizing a skill, agent, prompt, code path, benchmark, or other target with a mechanical metric: baseline, mutate one thing, verify, keep or discard, log, chart progress, and optionally save improved custom skills or agents through Open Cowork MCP tools.

Use the charts MCP to create clear, truthful, well-labeled data visualizations from structured data, including correct chart choice, chronological date axes, readable labels, and concise interpretation.

Design and author OpenCode skill bundles for Open Cowork, including SKILL.md and any supporting reference files.

Design and save repeatable Open Cowork workflows with manual, scheduled, or webhook triggers.

Use the time-keep MCP for reliable time context - IANA timezone ops, date arithmetic, calendar queries, business days, offline holiday lookups (2000-2030), and local timers.