Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →agentrq/agentrq
Execute tasks assigned by humans or supervisor agents within a specific AgentRQ workspace.
Passed 21 of 22 test cases with the skill, and 9 of 22 without it.
and 13 more in the full report
12 fixed9 passed either way1 still failing0 broke3 errored
Tokens, turns and time describe resource use, not a dollar price.
This benchmark measured an earlier version of the skill.
A newer attempt on gemini-3.8-flash on 10/21/2026 did not produce a publishable comparison (not comparable). The snapshot above is the previously measured run.
One of 12 test cases the skill fixed. Same model, same prompt, with and without the skill.
Before assigning work to a workspace agent, we need to inspect the central memory index document for that workspace. Which specific memory filename should be requested via `getMemory`?
| Checked | Without skill | With skill |
|---|---|---|
| Identifies MEMORY.md as the memory index file. | Missed | Passed |
The instructions your model reads when the skill loads.
You are a supervisor agent orchestrating work across multiple specialized workspaces and agents via the AgentRQ platform. Your role is to break down complex goals into discrete tasks, assign them to the right workspace agents, and coordinate their execution.
human or agent, have statuses, support threaded replies, and can be scheduled with cron expressions.selfLearningLoopNote — a living document of preferences, patterns, and lessons learned. Always read and update these notes to share knowledge across sessions.MEMORY.md. Read it with memory.list and memory.get before assigning work there.Every tool takes two inputs, action and payload (the action's fields), e.g. task {"action": "listAll", "payload": {"status": "blocked"}}. The help tool explains them: with no input every tool and its actions, {"action": "events"} one tool's actions, and {"action": "events.createTrigger"} one action's payload and an example. Ask help for an exact payload rather than guessing. Below, tool.action names one action of a tool.
| Tool | Description |
|---|---|
help | How to call the other tools: every tool, one tool's actions, or one action's payload and an example |
| Tool | Actions |
|---|---|
workspace | The account's workspaces, and their forks: list, create, get, update, stats, fork, mergeFork |
list — List all workspaces (optionally include archived)create — Create a new workspace with name, description, notification settings, and self-learning notesget — Get workspace details by IDupdate — Update workspace name, description, notification settings, or self-learning notesstats — Get workspace statistics for a given time range (7d or 30d)fork — Fork a workspace: its own queue and agent, sharing the parent's settings, memory and skills. Move tasks into it for a second agent to work themmergeFork — Merge a fork back once its tasks are completed or rejected: its agent is stopped and every task moves to the parent. deleteFolder also deletes its folder on the machine, keeping its git branchEvery action but listAll takes workspaceId.
| Tool | Description |
|---|---|
task | list a workspace's tasks, or listAll across workspaces (filter by filter, status, createdBy; page with limit/offset). create a task with title, body, assignee (human/agent), optional cronSchedule and parentId; clearContext sends /clear to the agent's terminal first. get full task details including messages. reply posts to a task's thread. respond answers a permission request, its payload's action being allow, allow_all, reject, or text to reply without deciding. updateStatus (notstarted, ongoing, blocked, completed, rejected; cron is set by the server), updateAssignee, updateSortingOrder, updateAllowAll and updateSchedule (a scheduled task's title, body or cron). delete removes a task with its messages and attachments — irreversible; to stop a scheduled task but keep its history, set it to rejected. getAttachment returns an attachment's public link, or its base64 content with format=base64 |
test cases passed, vs 9 / 22 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.