Writesonic Agents

View saved version →

---
name: agent-run-review
description: Walk through what an agent did on a task, step by step. Use when the user asks what an agent did, why a task looks the way it does, or to review agent work on a task.
---
Review agent work on task "$ARGUMENTS" (a task id or title; resolve with `query_tasks` if it's a title) in the user's project (`whoami` first).

1. `get_agent_runs` with the task_id — list the runs.
2. For the latest run (or the one the user names), show its progress steps in order.

Summarize: what the agent was asked, each step it took, the outcome, and anything that failed or was skipped. Read-only.

View saved version →

---
name: brand-voice
description: Answer a question from the Brand Kit or Knowledge Base, e.g. "/writesonic-agents:brand-voice what does our guide say about tone".
disable-model-invocation: true
---
`whoami` first. Answer "$ARGUMENTS" using `get_brand_kit` for voice/tone/positioning questions, falling back to `search_knowledge` across knowledge sources. Quote the relevant passages and name the source.

View saved version →

---
name: new-task
description: Create a task from a title, e.g. "/writesonic-agents:new-task Refresh the pricing page for AI search". Assigns it to you unless told otherwise.
disable-model-invocation: true
---
`whoami` first. Show the task you are about to create from "$ARGUMENTS" (title, project, assignee = the current user unless another is named) and ask for confirmation. On yes: `create_task`, then `update_task` with action="assign". Report the created task id.

View saved version →

---
name: status-snapshot
description: One-shot workspace status — open task counts, recent playbook runs, failing integrations. Use when the user asks for a status update, snapshot, or "how are things" across the workspace.
---
Give a status snapshot for the user's project (`whoami` first; use "$ARGUMENTS" as the project name if given).

Chain the reads in ONE `execute_batch` call rather than round-tripping:
1. `query_tasks` view="counts" grouped by status.
2. `get_workflow` view="runs" for the most recently used playbooks (from `list_workflows`).
3. `list_integrations` — flag any that are failing or disconnected.

Report as three short sections. Read-only.

View saved version →

---
name: tasks
description: List tasks matching a filter, e.g. "/writesonic-agents:tasks high priority open" or "/writesonic-agents:tasks assigned to me".
disable-model-invocation: true
---
`whoami` first. Interpret "$ARGUMENTS" as a filter (priority, status, assignee, keyword) and call `query_tasks` view="list" with the matching filters (default: open tasks). Show title, assignee, status, priority, due date. Read-only.

View saved version →

---
name: task-triage
description: Triage the task list — open task counts grouped by assignee, then the high-priority tasks that aren't done. Use when the user asks about open tasks, workload, who has what, or what's high priority.
---
Triage tasks in the user's Writesonic Agents project (call `whoami` first; if several projects, use the one named in "$ARGUMENTS" or ask).

1. `query_tasks` with view="counts", group_by="assignee_id" — open tasks per person.
2. `query_tasks` with view="list" filtered to high priority and not-done statuses — the list that needs attention.

Present counts as a table, then the high-priority list with title, assignee, status, due date. Read-only; do not modify tasks unless asked.

View saved version →

---
name: whoami
description: Show who you are signed in as and which Agents projects you can access.
disable-model-invocation: true
---
Call `whoami` and present identity, teams, and the projects list with ids. If "$ARGUMENTS" names a project, confirm it exists and use it for the rest of the conversation.